Plasma-facing part

During testing cracks occurred in the loaded surfaces due to thermally induced me- chanical stresses, too high compared to the strength of W materials (figure 2.7). The fracture mode was cracking by tensile forces (as described in section 1.7). The crack distance was in the range of the grain width (∼100 µm). The ground surfaces were eas- ily recognisable as they showed a smooth aspect even after thermal loads (figure 2.7(a)) whereas the as-machined surfaces appeared rougher (figure 2.7(b)). Thus, the grinding process did not prevent crack formation by thermal fatigue. In addition to cracking the components #14 and #32 (table 2.1) showed surface melting; the grains were molten individually (figure 2.7(c)). Surface melting was related to a defective bond between the

Figure 2.7. SEM micrographs of the loaded surfaces after (a) 18 cycles at

10 MW·m−2, (b) 89 cycles at 10 MW·m−2 and (c) 90 cycles at 9 MW·m−2showing

cracks due to thermal stresses. A few components showed surface melting on top of surface cracking due to a defective bond between the plasma-facing and heat sink parts.

plasma-facing and heat sink parts that led to partial or complete detachment of the tile from the thimble. Complete detachment led to significant crack opening (figure 2.8). The examination of the cross section of the loaded surfaces revealed that cracks induced by thermal fatigue formed dense crack networks in the plasma-facing part (figure 2.9).

Results 27

Figure 2.8. SEM micrographs of the loaded surfaces area after (a) 90 cycles

at 9 MW·m−2 and (b) 10 cycles at 10 MW·m−2 showing surface melting and

significant crack opening due to overheating of the plasma-facing part after its detachment from the heat sink part.

Figure 2.9. BSE micrograph of the cross section of the loaded area after 89

cycles at 10 MW·m−2. Cracks induced by thermal stresses in the plasma-facing

parts propagated and formed dense crack networks.

The crack depths in components #12, #13, #17 (table 2.1) were about 118, 163, 180 µm respectively. The crack depth showed a tendency to increase with the cycle

number among the components tested with a He MFR of 9 to 10 g·s−1 (figure 2.10).

The crack depths in components #5 and #21 (table 2.1) were shallower (43 and 69 µm respectively) although they were tested at higher cycle numbers (more than 100 cycles).

It indicated that the increase of the mass flow rate (from 9 to 13 g·s−1) led to a decrease

of the crack depth. The higher the mass flow rate, the greater the cooling efficiency. Consequently, the surface temperature during the thermal loading was lower and so were the thermal stresses. Therefore, the decrease of the operation temperature pre- vented the crack development to a certain extent. The crack depths of the components #14, #24 and #32 (table 2.1) were not measured because these components showed

a particularly severe failure mode related to joining issues between the plasma-facing and heat sink parts.

Among the components tested with a He MFR of 13 g·s−1, component #5 (with a tile

made of a Polema W rod) showed a lower crack depth compared to component #21 (with a tile from a Plansee W rod). The behaviour of the Polema W has to be rated even higher based on the fact the tile of component #5 had no castellation slots, implying greater thermal stresses, and was subjected to a higher cycle number than the tile of component #21. This difference in terms of crack depth and crack density showed that the resistance to crack propagation depended on the W grade.

Figure 2.10. Crack depth in the plasma-facing part as a function of the cycle number. Among the components tested with a helium mass flow rate (MFR) of 9

to 10 g·s−1, the crack depth showed a tendency to increase with the cycle number.

The increase of the MFR enhanced the cooling efficiency, reduced the thermal stresses and minimised the crack propagation in the loaded surface.

Due to differences in terms of testing parameters (cycle number, power density and mass flow rate) and material selection between the divertor modules, the impact of the surface finishing on the crack development could not be accurately determined.

The cracks induced by thermal stresses propagated along the grain boundaries. Thus, in the rods with a grain orientation parallel to the heat flux the cracks followed the direction of the grain orientation and propagated perpendicular to the loaded surface, towards the heat sink (figure 2.11(a)-(c)). The W plate used for the tile of the compo- nent #29 (table 2.1) surprisingly showed no particular grain orientation. Consequently, the cracks were remarkably shallow although the component was tested at a relatively

Results 29

Figure 2.11. BSE micrographs of the cross sections of the loaded surfaces after

(a) 89 cycles at 10 MW·m−2, (b) 100 cycles at 9.5 MW·m−2, (c)124 cycles at

9 MW·m−2, and (d) 39 cycles up to 14 MW·m−2. At the same helium mass flow

rate (13 g·s−1) Polema materials (rod and plate) showed better resistances to crack

propagation compared to Plansee materials.

The as-machined components #11 and #16 (table 2.1) showed discrete microcracks that occurred in the surfaces (contours and castellations) of the plasma-facing part during machining. These microcracks resulted from stresses due to the rapid heating and cooling effects induced by the discharges during the EDM process [127]. The crack depth was typically about 30 µm (figure 2.12).

Figure 2.12. BSE micrograph of the cross section of the as-machined plasma- facing part (e.g. component #11) showing microcracks induced in tungsten sur- faces by electric discharge machining (EDM).

As the grinding process removed 0.2 mm thick W layers, it also removed the discrete cracks induced by EDM except in the castellations that were too narrow (0.2 mm large) to allow any finishing tool to penetrate in-between (figure 2.13(a)). In the tested components, defects were observed on top of cracks in regions close to the castellations (figure 2.13(b)). Those defects could be related to pre-existing defects (e.g. cavities) in the raw W materials (i.e. rods and plates) that developed under thermal loads. It could also be related to machining-induced defects that developed by thermal stresses. Or the combination of both.

Figure 2.13. BSE micrographs of the cross sections of the tile (a) after machin-

ing and (b) after 70 cycles at 9 MW·m−2 showing microcracks in the castellations’

surfaces induced by EDM and cracks as well as defects in the same region respec- tively. The microcracks induced by EDM could have grown and futher damaged the components during thermal loads.