Tungsten as first wall material in fusion devices

doi:10.1016/J.FUSENGDES.2007.03.045

Tungsten as First Wall Material in Fusion Devices

M. Kaufmann, R.Neu

Max-Planck-Institut für Plasmaphysik Garching, EURATOM Association, GERMANY

Abstract

The observation in JET of co-deposition of tritium with carbon has led to a broad

discussion on the replacement of graphite by a high-Z material for the first wall coverage.

Moreover, due to the high erosion rate, carbon plasma facing components (PFCs) appear to

be unacceptable for a commercial fusion reactor. Research in this area has subsequently

gained increased attention. This paper describes the status of investigations on the use of

tungsten as a First Wall Material. It discusses on the physical side the plasma wall

interaction, the transport of tungsten in the plasma boundary and in the core. As an

intermediate step on the technological side, graphite is often coated with tungsten layers.

For highly loaded surfaces in a fusion reactor finally bulk tungsten components will have to

be developed.

Keywords

First wall materials, tungsten, plasma facing components, plasma wall interaction

1. Introduction

Strong cooling of the central plasma has been first observed in the PLT tokamak with a

tungsten limiter [1]. Since then graphite has mostly been used in tokamaks as limiter, or in

divertor configurations as target plate material. Only a few tokamaks (limiter: FTU [2],

TEXTOR [3]; divertor: Alcator C-Mod [4], ASDEX Upgrade [5, 6]) have gained

experience with high-Z materials. In machines only partly covered with tungsten a complex

migration of carbon from the main chamber into the divertor [7] and the other way around

[8] was found. With the observation in JET of strong co-deposition of tritium with carbon

[9] together with the results of design studies of fusion reactors [10], it has become clear

that in the long run tungsten will be the favoured choice for plasma facing components

(PFCs). In particular, the relatively short life time of low-Z material due to erosion is seen

as unacceptable in a fusion reactor. Among the high-Z materials, tungsten is the only one

with a relatively short activation decay time. Therefore, in the future, JET and ITER will at

least partly use tungsten as PFCs.

While for graphite a broad experience is available, tungsten requires intensive research in

several areas; specifically in plasma wall-interaction, in the physics of the confined plasma,

and in material development. In order to quantify the tungsten influxes and spatially resolve

the tungsten concentration a comprehensive spectroscopic diagnostic had to be developed

(see for example [11, 12, 13]).

In contrast to carbon, the erosion of tungsten by low energy hydrogen atoms is small.

However, in the case of tungsten, erosion by fast particles plays an important role.

Radiation by carbon in the plasma boundary reduces the load to the target plates and has to

be substituted in a pure tungsten machine by artificial impurity gas puffing. Tungsten in the

confined plasma is only partly ionized, even at reactor relevant temperatures in the range of

10 to 20keV. The resulting strong radiation thus sets a lower limit for an acceptable

tungsten concentration of only a few 10

-5

. In connection with the pronounced neoclassical

inward drift of high-Z ions, turbulent transport plays a beneficial role here. The inward drift

in the pedestal region of an ITER relevant H-mode with type –I ELMs [14] can lead to a

further increase.

Technical developments are mostly directed to the specific needs of existing devices where,

for example, graphite tiles are often coated with tungsten layers. In future devices like

ITER, the solid tungsten armour of the divertor target plates will have to be castellated

because of the difference in its thermal expansion compared to the cooling structure.

The plasma wall interaction, discussed in section 2, shows significant differences to carbon

– mainly due to the higher atomic mass. Section 3 deals with tungsten transport in the

confined plasma covering neoclassical and anomalous effects. A short description of

technical questions is given in section 4. The summary stresses the necessity for further

investigations.

2. Plasma Wall Interaction with Tungsten

One of the main advantages of tungsten is its low permanent hydrogen retention [15, 16] at

elevated temperatures. H-retention by co-deposition does not play a significant role with

tungsten due to the low expected erosion rates and the absence of a process similar to that

of hydro-carbon formation. The implications of the formation of H bubbles, as well as of

the blistering due to He, observed in laboratory experiments after irradiation with high

fluencies of H or He respectively, are still under discussion (see for example [17]).

Because of the higher atomic mass substantial sputtering by deuterium occurs only at

relatively high energies (e.g. sputtering rate 10

-3

for 400eV ions) so that sufficiently cold

deuterium plasma in the SOL and in front of the target plates does not lead to substantial

erosion. In addition, because of their low speed, W-neutrals sputtered from the wall are

ionized over distances small compared to their large gyro radius. This leads to a prompt

redeposition of a high fraction of sputtered particles [18].

In tokamaks with graphite surfaces, carbon plays an important role as a radiator in the

plasma boundary leading to a substantial reduced heat load on the target plates. With a first

wall of a high Z-material, the carbon radiator has to be replaced by Argon or Neon for

example [19, 20]. While with graphite surfaces, a degree of self regulation can be observed,

the addition of a noble gas has to be carefully controlled. To this end, the measurement of

currents through the target plates induced by thermal forces seems to be a robust method

for the control of the divertor electron temperature [19]. While the erosion of tungsten

surfaces by low temperature hydrogen can usually be neglected this is not the case for these

admixtures if the divertor plasma temperature is too high [11, 21, 22]. In a limiter

configuration, tungsten erosion is in any case unacceptably high while with a divertor the

plasma temperature in front of the target plates has to stay below about 5eV [23]. It requires

in the divertor configuration a strong reduction of the energy transport during ELMs

compared to that obtained with the standard type-I-ELMs. This can be achieved by external

means e.g. an increase in the ELM frequency [24] or by increasing the overall edge

transport with edge resonant magnetic field perturbations [25]. A specific effect in the main

chamber can be observed during ELMs. While the main fraction of the ELM energy loss

goes into the divertor, filaments with a helical structure [26] transport plasma to the wall of

the main chamber. Impurities in this typically T

i

≈ 50eV plasma [27] have the potential to

sputter tungsten. Erosion of the main chamber by CX-particles may also lead to substantial

tungsten fluxes if, in some confinement scenarios, the plasma edge temperature is relatively

high [20].

Another source for erosion are fast particles from additional heating. Fast deuterium ions

generated by Neutral Beam Injection can be lost due to drifts or MHD-effects, like ELMs.

Their effect on tungsten erosion of protection limiters has been observed with high time

resolution [28, 29]. In contrast, for the case of ICRH, tungsten erosion is not dominated by

fast particles [30, 28, 31]. Local observation on limiters shows tungsten influxes created

immediately after the ICRH is switched on, which suggests that an increase of the sheath

potential in the neighbourhood of the antenna is responsible. However, variations of the

field line angle with respect to the Faraday screen - which should lead to a strong variation

of the electric potential - do not show a clear effect. The limited theoretical understanding

poses a challenge for the combination of ICRH and a tungsten covered wall. The erosion of

tungsten by ICRH (and of course by other sources) can strongly be reduced by a

boronization of the surfaces, but boron disappears after a few discharges on heavy loaded

areas [32, 30, 33].

3. Edge and Core Transport

In the region of open field lines, the transport of tungsten has been described by the

DIVIMP-code based on a B2-EIRENE plasma background [34]. These results have been

extrapolated to ITER [35].

In the H-mode, one has to subdivide the confined plasma into the pedestal and core regions.

In the reactor relevant regime of type-I ELMs, the pedestal, with its steep pressure gradient,

breaks down during an ELM and a substantial part of the pedestal plasma is ejected. In

between ELMs, tungsten moves into the pedestal region due to strong inward particle drift

[14]. If the next ELM comes in due time then this tungsten is removed. One can clearly

observe this effect as a function of the ELM frequency (see figure 1) [24]. This suggests

that ELMs are a necessary prerequisite for sufficiently low tungsten content in the plasma.

For ITER or DEMO this demand is in line with the necessity to keep the energy per ELM

small (see above).

In the core plasma, however, an inward particle drift can lead to accumulation in the centre.

In the simple picture, neoclassical theory predicts a diffusion coefficient which scales

proportional to 1/Z

2

while the inward drift is proportional to 1/Z. If the deuterium density

profile is not particularly flat this, in general, leads to a strong accumulation of tungsten in

the core. The neoclassical accumulation has been experimentally observed in several

occasions [36, 6, 37, 38, 39]. However, if the heat flow in the core is sufficiently high then

anomalous transport can easily exceed the neoclassical effects, especially of high-Z ions

because of the 1/Z and 1/Z

2

scaling respectively. One can experimentally demonstrate this

effect by, for example localized central ECRH which stimulates the anomalous turbulent

transport [40, 6, 41] as demonstrated in figure 2. There, the top three rows depict the

trajectories of the additional heating (please note the small contribution of central ECRH)

and the temporal behaviour of the H factor, the H

α

light and the electron temperature at two

radial positions. The tungsten content drops with central heating and recovers without,

which can be deduced from the comparison of central and edge W concentration (c

W

)

shown in the lowest row. There are two ways the anomalous transport can influence the W-

profile. On the one hand it can overcome the Ware-pinch of the main ions, thus flattening

the deuterium density profile (second lowest row in fig.2), while on the other hand

anomalous transport can directly determine the tungsten flux. The discharge in figure 2

does not develop sawtooth activity, which in ordinary H-Mode discharges also expells the

central impurity content (see for example [14]). Recent theoretical work [43] shows two

dominant turbulent transport mechanisms for high Z-ions. Neither of them leads to a

substantial accumulation of tungsten with respect to the deuterium density. In summary,

this leads to the expectation that peaked W concentration profiles are unlikely to exist in the

type-I-ELMy H-Mode of a burning device. Combined with a high density, low temperature

edge plasma, an all tungsten device seems to be feasible [20], although the details of the

radial transport still have to be quantified.

4. Technological Developments

On the technical side, the critical issues are the mechanical properties of tungsten. Although

its strength is high, its tensile elongation at room temperature is almost zero, making it

brittle. The ductile to brittle transition temperature (DBTT) is far above room temperature,

depending on the details of manufacture. In order to improve the brittleness several kinds of

tungsten based alloys have been developed.

In most present day devices, inertial cooling is sufficient due to the short pulse durations.

The low particle fluencies, and therefore, the limited erosion, enable the use of W coatings

on a graphite substrate. In steady state devices, actively cooled components will be

necessary and for high heat flux components, the bonding between the W armour and the

heat sink is of special concern. In the case of ITER, copper alloys are proposed as the heat

sink material. Since copper exhibits a much larger thermal expansion than tungsten,

castellated or `brush'-like structures will have to be used. In order to operate at higher

cooling medium temperatures, concepts of actively cooled tungsten structures are under

development for DEMO.

Specific developments are aimed at suppressing the production of volatile tungsten oxide in

case of a loss-of-coolant accident with air ingress so as to reduce the risk of releasing

radioactive material. The admixture of small amounts of silicon and chromium to tungsten

leads to a glassy protection layer on the tungsten surface, which can reduce the tungsten

oxidation rate by up to four orders of magnitude (see Fig. 3) [44].

4.1 Tungsten Coatings

Graphite or carbon-fibre composites (CFC) have a reduced electrical conductivity

compared to metals which helps to keep eddy and halo currents low. In contrast, tungsten

has a conductivity 200 times as large as carbon based materials and its mass density is

larger by a factor of 8.5. These properties would lead to a considerable higher load on the

support structures if bulk tungsten tiles were used making the transition from a device

designed for C-based PFCs very costly and time consuming. In principle, castellation of the

W components could be used to reduce eddy currents, but the technical effort is quite high

(see for example [45]). Therefore, several present day devices have chosen the coating

solution, when implementing tungsten PFCs. A variety of techniques are available,

amongst which physical vapour deposition (PVD) [46, 47, 48] and plasma spray (PS) [46,

49] are most commonly used for fusion applications to produce thin (µm scale) or thick

(mm scale) coatings, respectively. They can withstand heat loads in excess of 10 MW/m²

with typical erosion rates below 1 nm/s [50] and they can survive 10

4

– 10

9

seconds of

plasma operation. Chemical vapour deposition (CVD) has been investigated as a method

for in-situ W-coating of PFCs in fusion devices [51]. Recently, JET has launched a test

program in the framework of the ‘ITER-like wall project’ to investigate PVD, CVD as well

as PS coatings on CFC substrate. Although the thermal expansion of CFC is strongly an-

isotropic and does not match that of tungsten, PVD and PS coatings could be identified in

thermal screening as well as cyclic loading experiments to serve the JET requirements [52].

Tungsten as first wall material in fusion devices

Summary (2 min read)

1. Introduction

2. Plasma Wall Interaction with Tungsten

3. Edge and Core Transport

4. Technological Developments

4.2 Bulk Tungsten Components

5. Summary

Figure Captions

Figures (3)

Citations

Cites background from "Tungsten as first wall material in ..."

References

Related Papers (5)