Deuterium retention in cyclic transient heat loaded tungsten with increasing cycle numbers

Surface damage and microscopic defect evolution of tungsten (W) armor under transient heat loads are key factors for fuel retention in fusion reactors. In this work, experiments were conducted to investigate the effects of cyclic thermal shocks on deuterium (D) retention and surface blistering in W. Thermal shock experiments were conducted on recrystallized W using an electron beam with a power density of 0.15 GW m−2 across 100–1500 cycles, followed by D plasma exposure with high-fluence (∼1 × 1026 D m−2). The results demonstrate that samples subjected to 500 and 1500 cycles exhibit a significant presence of sub-grains within 90 μm. Notably, the inhibition of blistering induced by thermal shock leads to a substantial reduction in D retention (5.45 × 1019 D m−2) at lower cycle numbers (100 cycles) compared to the reference sample (2.35 × 1020 D m−2) which was only exposed to D plasma. When cycle numbers increase to 500 and 1500, D retention reaches 1.98 × 1020 D m−2 and 4.56 × 1020 D m−2, respectively. Based on the tritium migration analysis program, we propose that total D retention is a consequence of the competition between defects reduced by thermal shock-induced suppression of blistering and defects generated by plastic deformation induced by thermal stress. D retention initially decreases with the increase in cycle numbers, followed by a subsequent rise, with the inflection point slightly higher than 500 cycles. Additionally, due to the extensive scope of thermal stress, an escalated exposure period will result in substantial D captured by heat-induced defects, consequently intensifying the D retention. Whether there exists an upper limit to D retention induced by the increasing thermal shock cycles necessitates further experimental analysis. Nonetheless, it is evident that thermal shock significantly contributes to D retention within a profoundly deep bulk region under high cycles.


Introduction
Fuel retention is one of the pivotal factors influencing the long-term stable operation of fusion reactors, which would adversely impact the thermal-mechanical properties of the material and raise concerns regarding safety and cost issues due to the radioactivity and rarity of tritium [1].As the divertor material in fusion reactors, tungsten (W) [2,3] will be subjected to continuous bombardment by steady-state high heat and energetic particles such as hydrogen isotopes (HIs) and helium (He).Moreover, W will also experience transient events [4][5][6][7][8], such as edge localized mode events (ELMs) [9], which are of particular concern due to the significant damage they cause and are inevitable in fusion devices.A large temperature gradient and high thermal stress will be induced by ELMs which power densities up to 1 GW m −2 for 0.2-0.5 ms [10].Therefore, it is crucial to comprehensively investigate the impact of transient thermal shocks (ELMs) and other associated factors on fuel retention in W under actual operating conditions.
Some studies have illustrated that both sequential and simultaneous exposure of heat loads and deuterium (D) plasma affect D retention in W [11,12].The synergy between H plasma and thermal shock significantly lowers the threshold for surface cracking in materials compared to thermal shock alone, accompanied by distinctive surface alterations [13][14][15][16].Experimental findings involving the performing of H blisters on W surfaces via hydrogen plasma irradiation, followed by thermal shock, further demonstrate a reduction in the fatigue resistance of W attributed to the blister structure [17,18].
As mentioned, numerous experiments have investigated the macroscopic surface damage behavior under the synergy of thermal shock and D plasma exposure.However, the current understanding of the microscopic mechanisms governing the impact of thermal shocks on D retention within W, as well as the comprehensive macroscopic alteration patterns, remains unclear.
The H retention primarily occurs due to the trapping by microscopic defects within W, such as vacancies, dislocations, grain boundaries and so on.In the case of dislocations, H exhibits a notable trend for diffusion into the dislocation core, facilitated by dislocations that serve as effective trapping centers for H atoms [19].For vacancies, they can strongly bind with HIs with the binding energy of ∼1.18 or 1.40 eV at distorted octahedral sites [20][21][22].The existence of vacancies introduces a substantial reservoir for H.Moreover, Von Toussaint et al [23] have illuminated the influence of grain boundaries on the dynamics of H trapping and diffusion within W using molecular dynamics.These defects comprise inherent defects within the material as well as those induced by plasma exposure and thermal shocks occurring during fusion reactions.Extensive studies [24][25][26] indicate that D plasma exposure induces the formation of blisters, accompanying a substantial presence of dislocation-type defects [27].Liu et al [24] proposed that blisters dominate D retention when blistering is severe.Besides, the high thermal stress induced by transient head load could result in plastic deformation on the surface of W, implying the multiplication and movement of dislocations.The related research [28] observed that the density of dislocations in W increases with an escalation in the number of thermal shock cycles.In addition, Yuan et al [29] have indicated that the roughness induced by thermal shock is mainly due to the shear bands whose directions aligned with the traces of {112} planes, which indicates a substantial activation of dislocation slip on the {112} planes during plastic deformation.
It is worth noting that the surface roughness induced by thermal shocks might play a role in inhibiting the formation of D-induced blisters.Manhard et al [30] have conducted experiments with W samples subjected to various surface treatments (mechanical grinding, electro-polishing, and chemical etching of varying durations) to obtain differing levels of surface roughness (0-1.02µm).Following D plasma irradiation, it was observed that an increase in surface roughness could reduce the formation of surface blistering.Therefore, the increased defects caused by thermal shocks and the inhibitory impact of thermally induced roughness on D-induced blistering might result in opposing impacts on D retention in W.Moreover, a critical question remains unanswered: whether a discernible threshold exists for the competitive results of surface roughening and microscopic defects on D retention.To achieve this objective, experiments involving different cycle numbers of thermal shock are necessary to obtain varied surface roughness and defect densities, ultimately unraveling the comprehensive impact of thermal shock on D retention.
This study aims to explore the influence of transient heat loads on D retention and surface blistering in W. In this study, recrystallized W samples are subjected to transient thermal shocks resembling ELMs, utilizing low power-density conditions of 0.15 GW m −2 , with pulse durations lasting 1 ms and varying cycle numbers ranging from 100 to 1500.Then these heat-damaged samples were exposed to D plasma.We investigated the effect of the cycle numbers of transient heat loads on the surface roughness and microstructure.Subsequently, we characterized and analyzed the D retention and surface blistering of W following plasma exposure with the change of macroscopic roughness and microstructure.The influence of transient heat loads on D retention in W is discussed based on thermal desorption spectrum analyses (TDS) and the tritium migration analysis program (TMAP).The correlations discovered between cycle numbers of thermal shock, surface roughness, hardness, surface blistering and D retention contribute to a deeper understanding of material responses to transient heat load.

Sample preparation
W samples with dimensions of 10 × 10 × 1 mm −3 were cut from a polycrystalline hot-rolled W sheet with purity >99.95%.The grains are elongated parallel to the surface.The samples underwent mechanical grinding using SiC sandpapers followed by electrochemical polishing with a sodium hydroxide solution to achieve a polished, mirror-like surface.Subsequently, they were cleaned in an acetone bath by ultrasonic cleaning and then the samples were annealed at 1973 K for 1 h in a high vacuum (below 3 × 10 −4 Pa) to accomplish full recrystallization.Figure 1 illustrates the surface morphology of the prepared W samples.

ELM-like thermal shock
ELM-like thermal shock tests were conducted using the 60 kW electron-beam material-test scenario (EMS-60) located at the Southwest Institute of Physics [31].The samples were bombarded with accelerated electrons at a voltage of 120 kV.The heat absorption power density was set to 0.15 GW m −2 .The focused electron beam was directed at a square area of 4 × 4 mm 2 near the center of each sample.Before loading, samples were heated to 500 K which closely approximates the ductile-to-brittle transition temperature of W [32]. Based on this temperature, the samples were subjected to 100, 500, and 1500 cycles of pulses with a duration of 1 ms and the period was set to 1 s to ensure the loaded surface cooled to base temperature [33].

Deuterium plasma exposure
Following the ELM-like thermal shock tests, the samples underwent exposure to high-flux D plasma using the linear plasma device STEP at Beihang University [34].The ion flux of ∼1.85 × 10 22 D m −2 s −1 was measured by a Langmuir probe.The incident ion energy was set as 40 eV by negatively biasing the sample, and the fluence was 1 × 10 26 D m −2 .The sample temperature was about 500 K as monitored by a K-type thermocouple attached to the rear of the sample.Additionally, a reference W sample was exclusively exposed to D plasma without undergoing prior thermal shock, designated as Ref.
Besides, the parameter details of heat loads and D plasma exposures are shown in table 1.

Post-exposure analysis
After the transient thermal shock, the surface roughness (Ra) of the samples was assessed using white-light interferometry.Five random areas of 0.8 × 0.8 mm 2 within the thermal loading region, were selected to calculate the roughness of the damaged area.The surface morphology observation of samples was conducted via scanning electron microscope (SEM, TESCAN MIRA 3 LMH).The orientation distribution of grains was examined by electron backscattering diffraction (EBSD, Oxford instrument Nordlys-Max2).
Hardness was evaluated by a Vickers hardness tester at room temperature.The test involved applying a load of 200 gf with a dwell time of 10 s, employing a rectangular pyramid diamond indenter.For each specimen, indentations were made on the grain surface with consistent grain orientation, ten indentations for {100}, {111} and 3-5 indentations for {110} orientated grains, determined by EBSD analysis.The lower indentations in {110} orientated grains are due to the {100} and {111} textures alongside a minimal amount of {110} grains in recrystallized W. Notably, due to the high number of thermal shock cycles in the 1500 cycles specimen, grain orientation was not discernible, and thus, indentations were randomly selected.To mitigate any potential influence from previous indentations, each indentation was placed on different grains, while avoiding contact with grain boundaries.The hardness values for specific grain orientations were averaged across the ten indentations.Vickers hardness number was automatically calculated by the following equation: VHN = 1.8544 × F D 2 where VHN is the Vickers hardness number in kgf mm −2 , F represents the load in kgf and D 2 denotes the area of the indentations in mm 2 [35].
D trapping and retention were characterized through TDS conducted on the HiTDS platform available at Beihang University.Before TDS measurements, the irradiated samples were kept in a vacuum at 1 × 10 −3 Pa for three months to ensure the release of interstitial D atoms.Samples were then heated from room temperature to 1273 K at a constant ramping rate of 1 K s −1 in a vacuum chamber.The W-25Re thermocouple affixed to the rear of the samples was utilized to monitor the sample temperature.During the TDS measurements, a sensitive quadrupole mass spectrometer (MKS Microvision Plus) monitored the signals of D 2 (mass 3) and HD (mass 4).The uncertainty of results was estimated to be about 7%.The combined quantity of released D from the exposed samples comprises both D2 and HD.
The TMAP code was utilized to simulate the results of the TDS, providing detailed insights into the D trapping energy and density distribution of defects.A comprehensive description of the TMAP code can be found in [36].TMAP has been extensively employed for simulating HI diffusion and desorption [24,26,37,38].In light of the rapid D release from the surface, the boundary condition prescribes a zero D concentration on the surface [37,39].TMAP operates under the assumption that each trap is occupied by a single atom.This single occupancy hypothesis suffices for describing H trapping in this work since the two hypotheses of single and multiple occupancies can only be distinguished in experimental isotope exchange at low temperature [40].Moreover, it is assumed that D atoms fully saturate the traps, meaning that the distribution of trapped D reflects the distribution of the traps themselves.
For accurate simulation of the 3D temperature and stress distribution in the heat loading area during pulsed loading, finite element methods (FEMs) were employed [29].ANSYS workbench version 2021 R2 is utilized as the FEMs simulation tool.ANSYS transient thermal simulation system was chosen for the analysis.To account for sample symmetry, onequarter of the full tungsten plate was used as the simulation model.Accordingly, the heat-loaded area was 2 × 2 mm 2 square area at the sample corner.Throughout the heat loading process, the heat flux was maintained at a constant value of 0.15 GW m −2 .Detailed information regarding the boundary conditions and thermal properties of recrystallized tungsten have been described in [33].The simulation results, providing temperature and stress distribution data, served as a reference for determining the required defect depth distribution for subsequent TMAP calculations.

Surface morphologies after thermal shock
Figure 2(a) illustrates the 3D temperature distribution of the model at the end (1 ms) of the heat loading, as obtained through FEM simulation.The simulation results demonstrate that the maximum temperature appearing at the center of the top surface of the samples is about 773 K, loaded at the base temperature of 500 K with 0.15 GW m −2 for 1 ms.Figures 2(b) and (c) show the 1D temperature and compressive stress distribution along the depth (Z-axis) direction of the model, ranging from the surface to a depth of 0.5 mm, as indicated by the white arrow in figure 2(a).From figures 2(b) and (c), it is observed that at a depth of 0.1 mm, the temperature is approximately 670 K and the stress value is around 300 MPa.According to relevant studies [41], the yield strength of tungsten at 673 K (400 • C) is slightly higher than 300 MPa.As we all know, the yield strength of tungsten will decrease with increasing temperature.Hence, under the low cycles (⩽1500 cycles) loading conditions of this experiment, it can be preliminarily inferred that the depth range notably affected by thermal stress should be within 0.1 mm.
Figure 3 presents the surface morphologies of the thermal loading region of recrystallized W samples after undergoing ELM-like thermal shocks with varying cycle numbers.SEM images at varying magnifications are displayed to depict the surface profile of the damaged region.In the 100 cycles, the surface is relatively flat (figure 3(a1)), with slight roughness observed due to the presence of parallel slip bands on each grain surface (figure 3(a2)), resulting in a subtly rough grain surface.For the 500 cycles, the surface roughness noticeably increases, accompanied by the emergence of different changes in grain surface morphology (figure 3(b1)).On some grain surfaces, fold-like convex features are notably pronounced, resulting in a visibly roughened surface.Other grain surfaces do not exhibit conspicuous wrinkle-like protrusions but rather display significant changes in contrast.These contrast alterations signify localized shifts in grain orientation, with areas of contrast differentiation forming numerous enclosed masses, suggesting the possible development of sub-grain structures (figure 3(b2)).In the case of 1500 cycles, the surface is further roughened, with prominent bulging observed along the grain boundaries.The presence of enclosed mass structures on the grain surfaces is notably more pronounced, both in terms of quantity and visibility.Moreover, a significant increase in prominent fold-like undulations has been observed on the grain surfaces compared with 500 cycles (figure 3(c2)).The plastic deformation of the heated grains can be attributed to the stress induced by the heat loading [42].The surface roughness of all investigated samples in the thermal loading region was depicted in figure 4, indicating a progressive increase in surface roughness with an increase in the number of loading cycles.
According to the backscattered electron (BSE) image shown in figure 5, a light contrast change was observed in 100 cycles (figure 5(b)).As the number of cycles increased to 500 and 1500, a notable development of sub-grain structures became evident (figures 5(c) and (d)).Further analysis   using EBSD demonstrates that these sub-grains exhibit an orientation deviation ranging from 2 • to 15 • (figures 6(c) and (d)).The proportion of low-angle grain boundaries (red line in figure 6) to all grain boundaries is 77.1% and 83.0% for 500 and 1500 cycles, respectively.Then the BSE image of the cross-section of 1500 cycles (figure 7(b)) indicates that subgrains are observed within a range of up to 90 µm in depth, consistent with the high-stress gradient zone estimated by FEMs simulation.These sub-grains primarily distribute within 30 µm in depth and exhibit a non-uniform distribution.The results obtained from BSE and EBSD analyses corroborate that the enclosed mass structures observed in the SEM results (figure 3) indeed represent sub-grains.Additionally, the predominance of low-angle grain boundaries in figure 6 suggests that the presence of these sub-grains may arise from dislocation slip within originally recrystallized grains due to thermal stress.This process leads to the formation of high-density dislocation walls, encompassing dislocation lines and tangles.Eventually dislocation walls evolve into sub-grain boundaries, due to continuous and coordination deformation [43][44][45].Consequently, with an increase in the number of thermal shock cycles, plastic deformation steadily accumulates, which will result in a significant increment in the number of dislocations, sub-grain structures, and other microstructural defects within the surface and subsurface regions.The formation of these defects may potentially change the mechanical properties of the material and affect the D retention during subsequent D plasma exposure.

Surface hardness
As illustrated in figure 8, the test result indicates that the hardness number of the samples increases with the cycles of thermal shock.Besides, no significant difference in grain orientation was observed in response to hardness changes.The hardness of {110} oriented grains slightly differs from the other two orientations.This discrepancy could be attributed to a statistical deviation resulting from the relatively fewer counts of {110} oriented grains, since the prominent presence of {100} and {111} textures in recrystallized W.However, this minor hardness deviation does not affect the overall trend.
It is imperative to note that although surface roughness (figures 3 and 4) and sub-grain structures (figure 6) showed only slight increases after 100 thermal shock cycles, dislocations formed in 100 cycles did contribute to material hardening.As cycles increased, the continuous energy input from thermal shocks facilitated dislocation multiplication, motion, and interaction, resulting in more pronounced plastic deformation and the emergence of numerous sub-grains.Notably, the substantial presence of sub-grain boundaries in the 500 and 1500 cycles could significantly hinder dislocation slip, further enhancing material hardening, and previous studies have shown that the cells and wall structures of dislocations will reduce the rate of work hardening [46].

Surface morphologies after deuterium plasma exposure
Figure 9 demonstrates the surface morphologies of recrystallized W samples after exposure to D plasma following transient thermal damage.Low-magnification SEM images are provided to depict the quantity and distribution of blisters.The measurement area for surface blister statistics was 4.62 × 10 5 µm 2 for each sample, and specific details can be found in table 2. The statistics on the surface blister encompass intergranular blisters and intragranular blisters which are the typical blistering types of recrystallized W [27,47].
For the Ref sample, the severity of surface blistering is the highest, with a number density of 2.47 × 10 −4 µm −2 and a surface coverage fraction of 1.69%.In the case of 100 cycles, the surface blistering is significantly mitigated, and the number density and surface coverage fraction of blisters are reduced to 8.45 × 10 −5 µm −2 and 0.48%, respectively.Notably, blisters are virtually absent on the thermal loading region of the 500 cycles and 1500 cycles samples.An important factor influencing the suppression of blistering behavior is the surface roughness induced by thermal shocks.Furthermore, as the roughness increases, this inhibitory effect becomes more pronounced.Other potential influencing factors will be addressed in the following discussion section.
Figure 10 displays the results of EBSD which is performed to investigate the relationship between blistering behavior and grain orientation.The measured results of the 100 cycles sample show that surface blistering still preferentially occurs on the grain with orientation close to {111}, which is consistent with the results of smooth surface samples [48].This suggests that surface roughness and microstructural defects induced by thermal loading do not affect the dependence of blistering on grain orientation.across all samples signify a uniform exposure temperature of 500 K.The desorption spectra exhibit two discernible peaks in all samples, which is also identified by the deconvolution with the Gaussian function.After pure D exposure, the Ref presents a small shoulder desorption peak at ∼600 K and a noticeable main peak at ∼720 K.The desorption peaks of the 100 cycles and 500 cycles show similar peak positions to that of Ref, appearing at ∼600 K and ∼700 K.While the intensity of the main desorption peak is significantly reduced in the 100 cycles sample compared to Ref, the 500 cycles sample exhibits a similar main desorption peak.In the case of the 1500 cycles sample, the desorption peaks show slight variations, including a main peak at around 700 K and a small shoulder peak at approximately 780 K. Notably, the main desorption peak in the 1500 cycles sample has the highest intensity among all investigated samples.

Deuterium desorption and retention
Besides, the total D retention is shown in figure 11, which is in the thermal loading region of all investigated samples after different numbers of thermal shock cycles while keeping As is well-known, the D desorption temperature is closely bound up with the defect types which function as D trapping sites in samples.The number of intrinsic defects in the initial samples would be very small since the W samples were recrystallized at 1973 K in this work.The previous study indicated a notably reduced dislocation density in W subjected to recrystallization at 2000 K compared to W subjected to stress relief treatment at 1200 K, which is (1.9 ± 1.4) × 10 12 m m −3 and (3.2 ± 1.7) × 10 14 m m −3 , respectively [49].Therefore, it can be inferred that defects induced by thermal shock and D plasma irradiation are likely the main factors contributing to significant D desorption.
Figure 12 presents the TMAP simulation results of the TDS for samples subjected to thermal shock and D plasma irradiation.It is noteworthy that the damage caused by    The less pronounced fitting of the shoulder peak at 600 K in figures 12(a) and (b) is a result of considering the overall fitting effectiveness.However, it does not impact the conclusions drawn from this work.The samples of Ref, 100 cycles and 500 cycles are characterized with two de-trapping energies of defects, defect 1 (1.17 ± 0.02 eV) and defect 2 (1.27 ± 0.01 eV).In addition, in the case of the 1500 cycles, defects of two de-trapping energies are identified as defect 2 (1.27 eV) and defect 3 (1.60 eV).These defects potentially correspond to grain boundaries (defect 1) [50], dislocations (defect 2), and vacancy (vacancy clusters) (defect 3) [51], respectively, and correspond to the ∼600 K, ∼700 K and 780 K desorption peaks in the TDS results.Figure 13 illustrates the distribution of D (defects) within each sample.The distributions of defect 2 exhibit a similar pattern across the 100 cycles, 500 cycles, and 1500 cycles samples.It is noteworthy that the concentration of defect 2 surpasses that of defect 1. Notably, in the depth range of 0-10 µm, the defect 2 concentration of 100 cycles, 500 cycles and 1500 cycles decreases successively, which are 1.77 × 10 −3 , 1.54 × 10 −3 and 1.20 × 10 −3 at.fr., respectively.In comparison, the defect 2 concentration of Ref is 1.24 × 10 −3 at.fr., lower than that of 500 and 1500 cycles but higher than 100 cycles.However, beyond the 10-20 µm depth range, the concentration of defect 2 experiences a sharp reduction in the 100 cycles (3.80 × 10 −4 at.fr.), 500 cycles (4.20 × 10 −4 at.fr.) and 1500 cycles (5.00 × 10 −4 at.fr.), reaching significantly lower levels than that of Ref (6.00 × 10 −4 at.fr.).Furthermore, the concentration of defect 3 in the 1500 cycles is notably high (1.10 × 10 −3 at.fr.) within the 0-2.5 µm range, yet it experiences a rapid decrease to negligible levels as depth increases.All these defects concentration experiences a rapid decline beyond a depth of 20 µm, attributed to the diffusion depth of D, and they decrease to 1 × 10 −5 at.fr. (the minimum concentration limit set in TMAP) with depth reaching 35 µm.Sun et al [52] calculated that under conditions of D flux at 4.40 × 10 22 D m −2 s −1 and temperature at 500 K, after 1.5 h of D diffusion, the concentration drastically decreased to a very low level (∼1.60 × 10 −8 at.fr.) by reaching 40 µm.Notably, the D concentration within the range of 0-20 µm was significantly higher than that beyond 20 µm.These parameters are closely akin to our experimental conditions (wherein the flux was ∼2.0 × 10 22 D m −2 s −1 , temperature was 500 K, and irradiation time was 1.5 h), and the simulation outcomes exhibit a notable alignment with our experimental findings.
We hypothesize that the concentration distribution of defect 2 (dislocations) in the 500 and 1500 cycles samples is a result of thermal stress, leading to the generation and evolution of a significant number of dislocation-type defects near the surface.The range of depth with a higher concentration of dislocations, as determined by fitting, aligns with the depth distribution of sub-grains (figure 7).The decrease  in dislocation concentration beyond 20 µm results from D absence due to the ability of D diffusion, thereby preventing its capture by deeper dislocations.Consequently, the utilization of desorbed D as an indirect descriptor of dislocation distribution in deeper regions becomes unfeasible.It is important to note that this phenomenon does not imply the absence of dislocations and sub-grain boundaries in deeper regions, although thermal stress does decrease as depth increases, and so does the density of dislocations.
For Ref, the distribution of dislocation defects is determined by D-induced blisters (intergranular and intragranular blisters), which is the reason that a high concentration remains at greater depths (the average depth of the first grain boundary is 20 µm).In the case of the 100 cycles, dislocations result from both thermal stress and D-induced blistering.However, due to lower levels of thermally-induced plastic deformation and blistering, the dislocation concentration is lowest.Besides, in the 1500 cycles, the presence of a substantial number of vacancy (vacancy cluster) defects could be attributed to intensified dislocation movement resulting from elevated plastic deformation in the near surface (0-2.5 µm), since the dislocation movements and interactions will lead to the formation of vacancies by two formation mechanisms [53,54].One mechanism involves the intersection of two dislocations with opposite Burgers vectors, leading to the formation of a row of vacancy-type defects.The other mechanism involves the movement of jogged dissociated dislocations.

Discussion
In this study, we conducted a comprehensive investigation into the effects of sequential transient heat loads and D plasma exposure on surface blistering and D retention in recrystallized W. Our research uncovered a significant correlation between the cycle numbers of transient thermal shocks and the occurrence of surface blistering as well as D retention.
One notable observation is the increase in surface roughness of W samples with higher cycle numbers of thermal shocks.The repetitive thermal stress leads to the formation of numerous defects and cumulative plastic deformation in the samples.Previous studies have also shown that thermal damage of W samples increases with an escalation in cycle numbers [55].Furthermore, the formation of sub-grains was observed in both the 500 and 1500 cycles, in line with the findings presented by Yuan et al [29].This study conducted a further analysis of the damaged region, revealing that subgrains predominantly reside within a depth range of 30 µm and have been observed up to 90 µm (figure 7).This result holds importance for subsequent analyses of D retention.
For the surface hardness, it exhibits an increase in the escalating cycles of thermal shock.However, with the increasing cycles of thermal shock, the predominant mechanism leading to material surface hardening could change.At lower cycles, the defects primarily obstructing dislocation slip are dislocation-type, resulting in an increase in hardness.In contrast, the substantial formation of sub-grains becomes prevalent at higher cycles, hence both sub-grain boundaries and dislocations could hinder dislocation slip under thermal shocks, leading to an increase in hardness.
Regarding the significant reduction and even complete disappearance of blisters with the increase in thermal shock cycles, we have discussed the pronounced inhibitory effect of surface roughness on blistering.Both the surface roughness induced by thermal shock in this work and the roughness obtained from chemical etching or mechanical grinding in the study of Manhard et al [30], demonstrate a suppressing effect on blistering.However, apart from roughness, there could still be other contributing factors influencing this inhibitory effect.In the experiment of Manhard, the mechanically ground samples (with surface residual stress) exhibited marked inhibition of blistering than surface stress-free samples, with a reduction of blister coverage by more than an order of magnitude.For this work, the surfaces of the 500 cycles sample reached a roughness of approximately 1 µm, similar to the roughness of mechanical grinding in their study.Moreover, the loading conditions during thermal shocks, including high strain rate, rapid temperature variation and so on, are expected to be more severe compared to mechanical grinding.Consequently, mechanical damage and residual tensile stresses parallel to the surface [56] on the W can be exacerbated under thermal shock conditions.Hence, the absence of blisters is reasonable in samples subjected to over 500 thermal shock cycles.
It is well known that during plastic deformation if the surface experiences residual tensile stress parallel to the surface, the interior of the material exhibits compressive stress.The increase in residual internal stress means that the nucleation and growth of blisters would have to overcome the additional internal pressure.It is plausible that residual tensile stresses near the surface stemming from thermal stress in this experiment could contribute to the absence of blistering.Moreover, the thermal-induced defects could be the barrier for the dislocation slip during the plastic deformation of blistering, which means that the hardening caused by thermal shock might inhibit the formation of blisters.Hence, the potential causes for the absence of blisters could be attributed to factors such as thermally induced surface roughness, microstructural defects, and surface residual stresses.However, the specific reasons require further experimental exploration and validation.
The TDS results indicate that D retention initially decreases and then increases with the increase in the cycles of thermal shock as shown in figure 11.As we all know, the generation of blisters will lead to the formation of numerous defects [24].In this work, the increase in the cycle numbers of thermal shock will lead to the obvious reduction or even disappearance of surface blistering.Consequently, the contribution of D retention caused by blistering is significantly reduced or eliminated.This is consistent with the result observed in the 100 cycles sample.In this case, the contribution of D retention from heat-induced defects does not compensate for the reduction in D retention resulting from mitigated blistering, leading to a lower total D retention compared to the reference sample (Ref).Additionally, the reduction in D retention after 100 cycles could also be attributed to an increased probability of reflection of the D plasma beam due to surface roughening.This leads to a decrease in the total number of solute D atoms entering the W.Moreover, the quantity of thermally induced defects may be insufficient to capture a large amount of solute D atoms.It is worth noting that D retention does not decrease significantly in 500 cycles and increases a lot in 1500 cycles compared with Ref.We attribute this to the increased D retention caused by defects (dislocations, sub-grain boundaries formed by dislocation movement and interaction, vacancies, vacancy clusters and so on) introduced through thermal shocks in our experiments.Consequently, the total D retention reflects a competition between defects reduced by thermal shock-induced suppression of blistering and defects generated induced by thermal shock.
This study revealed that the impact of thermal shock cycles on D retention follows a distinct tendency.As thermal cycles increase, D retention initially decreases, then increases again.The inflection point occurs when the defects introduced by thermal shock can precisely compensate for the reduction in D retention due to the absence of blisters.At the heat power density of 0.15 GW m −2 employed in this experiment, this inflection point is estimated to occur slightly above 500 cycles.It is reasonable to hypothesize that thermal shock experiments conducted at different power densities will also exhibit a similar impact trend, in which D retention initially decreases and then increases with escalating cycles, and the inflection point could appear at different cycles.Furthermore, the exposure duration of D plasma could also significantly affect the D retention under the synergistic effects of thermal shock and D plasma irradiation.This is due to the fact that the depth of defects induced by D-induced blistering in W will be noticeably smaller than the impact range of thermal shock (around hundreds of micrometers).Prolonged exposure will result in D diffusion into deeper layers, captured by thermal-induced defects.Hence, under experimental conditions of prolonged exposure, the increase in defects caused by thermal shock may predominantly influence D retention.Moreover, further experiments are required to determine if there exists a saturation limit for D retention as the number of thermal shock cycles continues to increase.Besides, the TMAP simulation results reveal that vacancy or vacancy-cluster-type defects are formed in the 1500 cycles sample, which can be explained by the motion and interaction mechanism of dislocations under high plastic deformation [53,54].This provides a reference to the types of defects formed in W under high-cycle thermal shocks for future fusion devices.

Conclusion
This study comprehensively explores the influence of sequential transient heat loads and D plasma exposure on D retention and surface blistering in tungsten.To achieve this, we meticulously designed samples with varying cycle numbers of thermal shock, which were subsequently subjected to D plasma irradiation.The final results indicate that the total D retention initially decreases and then increases with the increase in thermal shock cycles, which reflects a competition between defects reduced by thermal shock-induced suppression of blistering and defects generated induced by thermal shock.
Besides, the surface roughness of the W sample shows a noticeable escalation as the cycles increase.Meantime, the number of sub-grains is increasing with thermal shock cycles.Observations of cross-sections through BSE revealed that the sub-grain depth is up to 90 µm.Subsequent hardness testing unraveled a trend of increasing hardness with the rising cycles of thermal shock.The dominant factors contributing to this hardening are dislocations and the sub-grain boundaries, which could hinder dislocation slip during the plastic deformation stage.Moreover, surface roughness, residual stress, and defects, which are induced by thermal shock could be the main factors in inhibiting blistering.TMAP results indicate the emergence of vacancy or vacancy-clusters-type defects at high cycles (1500 cycles).
Despite the low cycle numbers of thermal shock in this experiment compared to those in actual fusion reactions, it is crucial and meaningful for exploring the entire process of thermal shock on D retention and blistering.Furthermore, the utilization of a low cycle numbers of thermal shock could offer insights into mitigating D retention in plasma-facing materials, thereby contributing to the development of radiation-resistant materials.For instance, this approach could involve introducing minor surface roughness as a pre-treatment strategy.By delving into the effects of transient thermal shock on blistering and D retention following D plasma exposure, we not only shed light on the underlying microscopic mechanisms responsible for the variations in D retention induced by heat load and D plasma synergism but also lay the groundwork for better comprehending the potential implications of ELM events on fuel retention behavior under fusion conditions in future investigations.

Figure 1 .
Figure 1.Scanning electron microscope image of as-received W after recrystallization at 1973 K for 1 h.

Figure 2 .
Figure 2. Thermal analysis result of the sample loaded with 0.15 GW m −2 for 1 ms (a) 3D temperature distribution of the model at the end of loading (1 ms); (b) the temperature distribution along the depth direction, as indicated by the white arrow in (a); (c) the stress (von-Mises) distribution along the depth direction, as indicated by the white arrow in (a).

Figure 3 .
Figure 3. SEM images of thermally induced modifications of recrystallization W samples after 0.15 GW m −2 thermal shocks at 500 K with (a) 100, (b) 500 and (c) 1500 cycles.The scale bar in each row of SEM images remains consistent.

Figure 4 .
Figure 4. Surface roughness of thermal loading region of investigated samples after transient thermal shocks.

Figure 11
Figure 11  depicts the TDS results of the samples, where the consistent initial temperatures of D desorption around ∼500 K

Figure 5 .
Figure 5. BSE images of the surface microstructures for recrystallized W after 0.15 GW m −2 thermal shocks at 500 K with (a) 0 (Ref), (b) 100, (c) 500 and (d) 1500 cycles.The scale bar of BSE images remains consistent.

Figure 7 .
Figure 7. BSE images of a local cross-section of the (a) Ref and (b) thermal loading area in 1500 cycles.The white dotted lines represent the surface of the sample.The scale bar of BSE images remains consistent.

Figure 8 .
Figure 8. Vickers hardness number of certain grain orientations variation with cycles of thermal shocks.
the parameters of D plasma exposure constant.The desorption signals of D2 and HD are both considered to acquire the total D retention.The D retention in the thermal loading region (4 × 4 mm 2 ) is calculated by subtracting the D retention in the undamaged region from the total retention in the whole sample (10 × 10 mm 2 ).The D retention in the undamaged region is obtained by multiplying the D retention per unit area in the Ref sample by the area of the undamaged region since the plasma flux is pretty uniform.The D retention in Ref is acquired by multiplying the same area of the thermal damage region (4 × 4 mm 2 ) by the D retention per unit area in Ref as reference.The results indicate that the D retention in 100 cycles is significantly lower than that in Ref.As shown in figure 11, the D retention in the thermal damage region prominently increases with cycle numbers.For 500 cycles, the D retention is slightly lower than that in Ref, 1.98 × 10 20 D m −2 and 2.35 × 10 20 D m −2 respectively.In the case of 1500 cycles, the D retention reaches the maximum (of 4.56 × 10 20 D m −2 ), which is 8.31 times and 2.3 times that in 100 cycles and 500 cycles, respectively.It is worth noting that the comparison of total D retention over the entire surface of all samples exhibits the same trend as that of the damaged area.Hence, to provide a clearer illustration of the influence of thermal shock on D retention, only the D retention comparison of the damaged area is presented.

Figure 9 .
Figure 9. Surface morphologies of thermal loading region of investigated samples after D plasma exposure.The scale bar of SEM images remains consistent.

Figure 10 .
Figure 10.Surface morphology (a) and surface normal direction IPF map (b) of 100 cycles.

Figure 11 .
Figure 11.Thermal desorption spectra of the samples which suffered different cycle numbers of thermal shock: (a) Ref (0 cycles), (b) 100 cycles, (c) 500 cycles and (d) 1500 cycles, and then exposed to D plasma at 500 K for 1.5 h.The desorption peaks are marked by gray dotted lines.

Table 1 .
Main parameters of heat loads and deuterium plasma exposures.

Table 2 .
The statistical data of surface blisters.