Understanding the reason of polycrystalline mirror degradation gives a clear hint that the step structure development can be possibly eliminated by applying mirrors from single crystals (SC). Experiments were performed with SC mirrors made from molybdenum, tungsten, nickel, and stainless steel with different orientations of the basic crystallographic axes (100), (111), (110). The samples were sputtered by deuterium plasma ions having a wide energy distribution. The behavior of SC mirrors was investigated in the course of long-term sputtering [3, 17, 18], during which the mirrors showed a considerably higher resistance to relief development under erosion than the polycrystalline mirrors did (Fig. 7). This result was corroborated by co-operative experiments at the large operating fusion facilities Tore Supra (France) [19] and TEXTOR (Germany) [20].

Based on these data, the ITER community has made the decision that SC molybdenum should be the first mirror material for plasma diagnostics in ITER. This material has a rather low sputtering coefficient and a moderate reflection coefficient R in the near ultraviolet and in the visible-light spectrum.

The above mentioned results indicate that the solution of the first mirror problem could rely only on the choice of those materials which exhibit a homogeneous surface erosion during long-term sputtering. As was mentioned above, such materials are metals that have a low sputtering rate (W, Mo) and a single-crystal structure.

Fig. (7). Comparison between the reflectances as function of the eroded layer thickness, for mirrors made from different PC and SC metals, and also, for a film mirror with a Rh film on a Cu substrate. The symbol “W (111) block” shows the data obtained at sputtering of PC textured tungsten with the (111) orientation of the majority (≈96%) of all grains [3, 18].

However, under ITER conditions there is a high probability that under impact of γ- and neutron radiation the single crystal will loose its perfect crystallographic structure, and then, the sputtering by charge exchange atoms may result in the development of surface roughness and the subsequent loss of optical properties. This was confirmed by experimental data on sputtering of SC Mo mirrors that have a defected structure, which showed a substantial degradation of the reflectance versus the sputtered layer thickness, though still at a lower rate in comparison to the degradation of PC Mo mirrors [3, 21]. To our knowledge, no experiments on single-crystal mirror surfaces have been performed in which the impacts of gamma and neutron radiation, and of ion bombardment are combined.

The production of single crystal mirrors with the required dimensions for ITER (≥10 cm) is not an easy task. Therefore, consideration has been given to another possibility of avoiding the appearance of the step structure at polycrystalline mirror sputtering, namely, through coating of the mirror surface with a metal film that has a rather high R (Rh, Mo) and thick enough (a few μm) to withstand the long-term sputtering by charge exchange atoms at the mirror location in the ITER port plug. Generally, such films have a fine-crystalline structure (≤100 nm). Therefore, it was expected [22] that the scale of microrelief, which appears after long-term sputtering may be appreciably shorter than the light wavelength range for which the mirror is applied.

Experiments with different films on various substrates (Be/Cu, Cu/Cu, Rh/Cu, Rh/V, Rh/SS, Mo/SS, Mo/Mo) have shown large differences in their behavior on exposure to deuterium plasma ions. If the film adhesion to the substrate is strong, the optical properties of these mirrors were almost not affected by long-term sputtering. For example, some Rh/Cu specimens prepared by pressing the films onto the Cu substrates, retained their properties even after sputtering of a layer of more than 7 μm thick (see Fig. 7) [3, 17]. Unfortunately, that technology cannot be applied (yet) for producing such thick film mirrors (thickness (10 μm) with an area large enough (~10 cm in diameter) to be of interest for application in ITER.

At present, at least three groups are investigating the prospect to use metal film mirrors for ITER plasma diagnostics: the University of Basel (Switzerland), the Politecnico di Milano (Italy), and the Indian Institute of Technology Guwahati (India). The Basel group operates with Rh films (and to some extent also Mo films) produced by magnetron sputtering and has a typical thickness ~2 μm (e.g., [23]). The two other groups study the behavior of films obtained by a pulsed laser deposition technique in vacuum [24, 25]. All main publications connected with investigation of film mirrors were recently enumerated in a short review [26].

Film mirrors were tested in small laboratory set-ups [27-29] and in large fusion devices TEXTOR [23], DIII-D [30], JET [31, 32]. Thanks to the improving technology there are prospects for such mirrors to be used in ITER. Note that because of the small film thickness and the rather high sputtering rate (for Rh film) the important condition for their use as first mirrors is to achieve a significant attenuation of the CXA flux density to the mirror surface, e.g., by strongly limiting the size of the input pupil. This was suggested for the ITER H_α monitor [33] and recently realized in experiments at T-10 [34] and JET [26] in which large size (200x300 mm) Rh film mirrors were employed.

Since the metal films are generally polycrystalline in their structure, at long-term sputtering their surface will gradually become rough with the result that the optical properties slowly degrade. This effect does strongly depend on the technology used for the film deposition. Some examples of this behavior, descibed in ref [35], were used in the analysis [11] of the experimental data obtained with polycrystalline mirrors. More details on the topic will be given below.

Mirrors made from amorphous metal alloys (AMA) may be another alternative to SC mirrors. Unlike the perfect structure of single crystals, the AMA have no structured order over a distance of several nanometers. Therefore the initial quality of their surface should remain the same, irrespective of the sputtering time. This assumption made as early as in ref [22], could be corroborated only after development of the technology for producing amorphous castings of sizes sufficient for preparing amorphous mirror samples (≥10 mm).

The experiments on mirrors made from zirconium- and titanium-based amorphous alloys have demonstrated their extra-high resistance to long-term sputtering [36, 37]: the amorphous mirror with composition Zr(41.2)Ti(13.8) Cu(12.5)Ni(10)Be(22.5) has kept its optical properties even after sputtering of a 13.4 μm thick layer (Fig. 8). The microscopic examination of the mirror surface has revealed no changes in comparison to the initial value before sputtering.

Fig. (8). Amorphous mirror reflectance at the given wavelengths versus the layer thickness sputtered for many exposures by ions of argon plasma with different fixed energies, ranging from 0.1 to 1.35 keV [36].

Although this sounds promising, AMA mirrors have the disadvantage that they absorb hydrogenic ions at a rather high rate. Certainly when operating ITER with tritium this would lead to a high tritium retention. Experiments on the interaction of a deuterium plasma ions of a few tens of eV with amorphous ZrTiNiCuBe mirrors were performed to study this effect [37]. Deuterium retention in these conditions is much more efficient than in experiments where hydrogen saturation was studied by exposing the AMA mirror to the gas phase or electrolytic solution. From the measured dependence of the absorbed deuterium mass on the ion fluence (up to 1.5·10²⁵ ion/m²) it was possible to estimate the efficiency of deuterium absorption to be ~10% for an ion energy of ~60 eV. This is about ten times higher than in the electrolytic hydrogenation experiments [38]. This is attributed to the fact that the hydrogen (deuterium) plasma ions of these energies can easily penetrate to a depth greater than the oxide layer thickness (a few nm), and can immediately get into the bulk of the material, whereas in the case of gas-phase hydrogenation, for example, the absorption process goes through several stages, viz., physical molecular absorption, molecular dissociation, chemical desorption, and only then inward diffusion follows [39].

The high absorptive capacity in relation to hydrogen isotopes holds no promise for the ZrTiNiCuBe amorphous alloys to be used as a material for the plasma facing mirrors in ITER. However, the rather rapid progress in the development of amorphous materials gives hope that in the future AMA mirrors with a low absorptive capacity (for hydrogen and its isotopes) will be designed, which would permit their use as a material for in-vessel mirrors under conditions of high fluxes of charge exchange atoms, neutrons and gamma-radiation.

Beryllium has been chosen as the first wall material of ITER, because of its low atomic number (Z=4) on the one hand in combination with the not too high a sputtering rate of Be under ITER conditions [1]. The choice of Be ensures a much lower contamination of the ITER plasma in comparison to any other metal wall, which need not necessarily be the case in future devices like DEMO. When the reactor is in operation, owing to erosion resulting from the plasma action, beryllium will propagate throughout the chamber, depositing at the sites most distant from the plasma confinement volume, including on the mirrors of optical and laser plasma diagnostics. In other words, there is a high probability that any metal mirror would become gradually covered with a beryllium film, and thus would change its optical properties. At the same time, in case of a Be mirror, the deposition of Be film on top is not expected to change much the optical characteristics of the mirror.

The basis for this assumption are data on the optical properties of beryllium that were analyzed in the paper “Beryllium” published in the Handbook of Optical Constants of Solids II, edited by Palik [40]. The majority of reflectivity data presented there were obtained with Be film deposited in vacuum. In the near UV and visible spectral range the film reflectance was noticeably higher than that of polished bulk beryllium mirror specimens.

It should be mentioned here that large differences have been observed between the optical characteristics of Be film deposited in a high vacuum [40] and in experiments with deuterium pressures of 0.5–0.8 Pa in “the PISCES-B divertor-plasma simulator” [41]. Analyzing the application of their results to ITER conditions, the authors of [41] wrote in the last conclusive remark that “…the lower pressure expected in the main chamber may lead to the formation of more dense layers, whose reflectivity is expected to be closer to the theoretical value of beryllium.”

In a detailed study on the behavior of Be mirrors exposed to deuterium plasma ions [42, 43] it was found that their optical properties strongly depend on the ion energy: with keV energy-range ions a significant drop of reflectance occurs already at relatively low ion fluence (≥5·10²¹ ion/m²). After the drop, the reflectance can be fully restored during a much longer time (ion fluence ≥1·10²⁴ ion/m²) by exposing the mirror to low energy ions (~50 eV) of the same plasma. For some Be mirror specimens similar procedures (drop and restoration of reflectance) were repeated several times without decrease of the level of reflectance restoration.

Fig. (9) illustrates the repetition of the process of reflectance decrease as a Be mirror was exposed for a short time to deuterium plasma ions of keV energy, and also, the reflectance recovery during a substantially longer time, when the ion energy is low (about 50 eV). Such reflectance behavior unambiguously indicates that all variations in the optical properties are determined by the chemical processes on the Be mirror surface, and the mirror surface roughness is likely to remain at the initial level.

The given data have provided the explanation of the behavior of Be mirrors, observed on their exposure to deuterium plasma ions of different energies. Each short-term action of keV D⁺ plasma ions under conditions of the experimental test bench DSM-2 (with some amount of water vapor) resulted in a gradual increase in the oxidized layer thickness. This process mainly proceeds through the formation of the hydroxide Be(OD)₂ by the reaction: 2BeO + 2D → Be(OD)₂ + Be. The first stage of this process is characterized by a complete transformation of the existing oxide film (extinction coefficient k ≈ 0) into a hydroxide film (here k > 0 [43]) that leads to the reflectance drop after exposure to a very low ion fluence of keV energy range, to a level of ≥5x10²¹ m^-2. The second stage is characterized by a relatively slower increase in the thickness of the oxide-hydroxide (possibly, only hydroxide) layer due to the reactions of free Be atoms either with new oxygen molecules to produce oxide, or with water molecules to produce hydroxide. This stage is accompanied by a gradual saturation of the reflectance drop at ion fluences of ≈2·10²³ m^-2 [43].

Fig. (9). Be mirror reflectance at the indicated wavelengths. 1 – reflectance drop due to 1350 eV ion bombardment; 2- reflectance recovery after long-term exposures to 60 eV ions; 3 – partial reflectance recovery after vacuum annealing for 2 hours at 200°C [43]. X1, X2, and X3 indicate the states of experiment when XPS measurements were provided (see Fig. 10).

Fig. (10). XPS data for the Be mirror exposed to deuterium plasma ions: ♦ – Ar+ion cleaning, Ei = 300 eV (X1 in Fig. 9); ● – reflectance drop after D+ ion bombardment, Ei= 1350 eV (X2 in Fig. 9); ■ – reflectance recovery after exposure to D+ions, Ei=60 eV, after the second reflectance drop (X3 in Fig. 9) [43].

As chemical analysis of the Be-mirror near-surface layer has shown (by the use of X-ray photoelectron spectroscopy, XPS), it is just this process that has been observed in our experiments with Be mirrors exposed to deuterium plasma contaminated with oxygen: after exposure of the Be sample to keV deuterium plasma ions, the oxidized surface layer became substantially thicker than the original layer (Fig. 10). At the same time, a long-term bombardment by low-energy D⁺ ions (E_i= 60 eV) leads to practically full recovery of both the near-surface layer thickness and the reflectance (Fig. 9), probably due to elimination of the thick oxide layer (see the discussion below).

An important feature of the second phase of the oxide-hydroxide layer thickness increase is the rate of free Be atom supply to the surface of the layer, because neither the oxygen molecules nor the water molecules can pass through this layer. The Be atom supply can be provided due to sputtering of the beryllium surface by deuterium ions that have passed through the oxide film: the higher the D⁺ ion energy, the higher the sputtering rate of the metallic beryllium and the higher the rate of oxide-hydroxide film thickness increase, i.e., the greater the reflectance drop [42, 43], Fig. (11). The saturation of the reflectance drop is determined by the fact that for each ion energy there exists a maximum path length in the oxygen-containing film.

Fig. (11). Reflectances of two Be mirrors at λ=220 nm as function of ion energy in the phase of reflectance degradation (black rhombs) from the initial (or partially recovered) state, and in the phase of recovery (red circles) after the degradation due to exposure to higher-energy ions [43].

To support the process described in this paragraph we can refer to ref [44], where the rate of the oxide layer thickness growth on the Be surface was one order magnitude higher when it was bombarded by 5 keV He⁺ ions than in the case of 2.5 keV D⁺ ions, due to probably a significant difference in Be sputtering yields for these projectiles.

If in the process of exposure, the ion energy gets reduced to a value where the ion path length becomes shorter than the oxide-hydroxide film thickness, the “source” of free Be atoms (i.e., the sputtering of metal bulk) will be terminated, and another process may start. This process has the opposite effect compared to the one which has led to the increase in the oxide-hydroxide film thickness, namely, the decrease in the oxidized layer thickness and the recovery of metallic beryllium occurs by the reaction BeO + 2D → D₂O + Be or Be(OD)₂ + 2D→2D₂O + Be. In Fig. (11), this “inverse” process is shown by red open circles. The rate of metallic beryllium recovery, i.e., the reflectance return to its initial level, goes at a considerably lower rate than the reflectance degradation, cf. the data shown in Fig. (9).

An important result obtained in Be mirror experiments is that two pc Be mirrors could withstand many procedures of reflectance drop and partial or full recovery (see Fig. 11) at the total ion fluence of 2.5x10²⁵ ions/m² with an average ion current density of 15 A/m² (the total time of exposure to the deuterium plasma exceeded 60 hours). By that time both Be samples still had a mirror-like surface and retained their high ability to transmit an image of a subject (the laser beam spot) practically with similar quality as an aluminum etalon did, refs [43, 45].

The results in this section proof that the reflectance variations are not connected to the development of Be surface microrelief, but are fully due to the chemical processes in the oxide film.

Qualitatively, the same effect as in Be mirrors was observed in exposure of Al mirrors to similar conditions. This is attributed to the fact that Be and Al are neighbors in the periodic table, and therefore there exists the so-called diagonal analogy, i.e., similarity in many chemical and physical properties. Our experiments have first established a close analogy in the behavior of Be and Al mirrors, as they are exposed to hydrogen plasma. This is apparent from the comparison between the data presented in Figs. (9) [43] and Fig. (12) [46].

The Al mirror experiments have provided additional insight into the processes decribed above for the Be mirrors. Secondary ion mass spectrometry (SIMS) and Auger spectroscopy techniques were used to perform layer-by-layer analysis of the near-surface layer of Al mirror samples after different operational procedures with exposure to deuterium plasma ions [46]. It has been found that a short-term bombardment by 1350 eV ions leads to an increase in the thickness of the oxygen-containing layer, while a longer exposure to low-energy ions of the same plasma causes an appreciable decrease in the layer thickness.

Fig. (12). Reflectance of an Al mirror sample at the indicated wavelengths versus the total ion fluence. Sharp falls are the result of short-term bombardment by 1350 eV deuterium plasma ions (stages 1, 2, 4) and with a wide energy distribution (100-1350 eV, stage 3). All rises are due to long-term exposures to low-energy ions (60 eV) [46].

This is testified, in particular, by an increased yield of AlO^- and AlOD⁺ ions (presumably, fragments of the molecules Al₂O₃ and Al(OD)₃, respectively) from the near-surface layer after a short-term exposure to 1350 eV ions, and by a substantial decrease in the yield of these fragments after a further long-time exposure of the samples to 60 eV ions, as the data of Fig. (13) demonstrate.

Fig. (13). SIMS analytic data on the composition of the near-surface layer of Al mirrors. The yields of AlO- and AlOD+ ions are given as functions of the depth of their occurrence [46].

The given depth distributions are very similar to the data for the ions of mass 2. These may be both the atomic ions D⁺ and molecular ions H₂⁺. However, since deuterium is the operating gas, it can be stated with high probability that a great part of biatomic ion current is related to deuterium ions (Fig. 8 in ref [46]). So, actually, these data qualitatively repeat the data obtained for the fragments of Al₂O₃ and Al(OD)₃ molecules.

The most demonstrative data have been obtained from the measurements of Al₃⁺ cluster yields from the same aluminium samples (Fig. 14). These clusters are easily formed during sputtering of a clean metal surface. However, if the surface is coated with an oxide film, the probability of cluster formation sharply decreases, while the yield of oxygen-containing fragments increases. In our experiments, after a short-term exposure to 1350 eV ions, the thickness of the layer, which showed a gradual increase in the ion cluster yield, considerably increased. However, after a long-term exposure to 60 eV ions, the transition layer became very thin again.

Fig. (14). SIMS data on the Al3+ cluster yield from the Al surface [43].

The SIMS data were confirmed by Auger spectroscopy as shown in Fig. (15). As can be seen, the oxide film thickness increases after bombardment by 1350 eV deuterium plasma ions, and substantially decreases after exposure to low-energy ions.

Fig. (15). Auger analysis data for Al mirror samples similar to those shown in Figs. (12-14). The solid lines (with open symbols) show the concentrations of the indicated components of the uppermost surface layer material as function of depth after exposure to deuterium plasma ions with energy 1350 eV, and the dashed lines (with closed symbols) show the same, but after a long time exposure to 60 eV ions. Initial data, after cleaning with Ar ions, are not shown for simplicity [46].

This section presents a more detailed description of the behavior of amorphous mirrors versus the conditions of their exposure to a deuterium plasma, using mainly the data of ref [37] as basis.

Studies were made in mirror samples of two types, containing the same components but with different atomic ratios: type 1 - from the Zr(41.2)Ti(13.8)Cu(12.5)Ni(10) Be(22.5) alloy manufactured at the NSC KIPT, and type 2 - from the Zr(46.75)Ti(8.25)Cu(7.5)Ni(10)Be(27.5) alloy developed at the Hahn-Meitner Institute (Berlin, Germany). Both alloys contained beryllium; therefore, their behavior could be expected to be similar to that shown by Be and Al mirrors. The performed measurements have confirmed this assumption.

The exposure of the mirrors to deuterium plasma ions of high and low energies has led to similar results as those obtained in the cases of Be and Al mirrors. This similarity, at least, qualitative, can be exemplified by the data given in Fig. (16) for the samples made from the Zr(41.2)Ti(13.8)Cu (12.5)Ni(10)Be(22.5) alloy. However, unlike the Be mirrors, the reflectance changes as a function of ion energy are considerably smaller; this probably is due to the relatively moderate content of beryllium in this alloy.

Fig. (16). The reflectance of an amorphous mirror specimen after exposure to 60 eV (solid circles) and 1.35 keV (open squares) ions from deuterium plasma. For this particular specimen the drop of reflectance was repeated three times and restoration after long-term exposure to low energy ions from the same deuterium plasma was repeated twice. In the figure the results after the last two procedures are shown.

Additional important information about the effect of the ion energy on the special features of deuterium plasma interaction with amorphous mirror surfaces has been collected with the SIMS method. In order to “sponge the memory” of previous exposures, all the samples before their exposure to the deuterium plasma were simultaneously bombarded by 1000 eV argon ions to a fluence sufficient to sputter a ~ 2 μm thick layer. Some of the data obtained are presented in Fig. (17). Each sample underwent the SIMS analysis in two points; the results obtained from both of them are in good agreement, therefore in Fig. (17) the results for only one point are presented.

The SIMS data show that after exposure to deuterium plasma ions of energy 1350 eV a noticeable increase in the layer thickness is observed for oxides of all metals that enter into the composition of the amorphous alloy. However, after exposure to 60 eV ions, the oxide layer thickness nearly regains the value, which was measured immediately after argon ion sputtering.

The fact that it is beryllium that plays a key role in the processes of amorphous mirror surface oxidation is in qualitative agreement with the results of paper [47], where the photoelectron spectroscopy technique was used for the analysis of near-surface layers of the same-composition samples.

The presented data show a qualitatively similar behavior of the reflectance of beryllium mirror and that of mirrors made from Be-containing amorphous metal alloys (Figs. 9-11, and Fig. 17).

The reflectance behavior of W and Mo mirrors under the action of deuterium plasma ions is somewhat similar to the behavior of Al and Be mirrors, i.e. the reflectance decreases after a rather short-term exposure to ions in the keV energy range [48]. The value of reflectance decrease, ΔR(λ), depends on the wavelength, viz. it decreases non-monotonically for the Mo mirror, and – rather monotonically in case of the W mirror. Fig. (18) shows the data obtained in experiments with four SC Mo specimens (a) and two SC W specimens (b). As seen, there is a quite reasonable qualitative agreement in values of initial drop of reflectance for Mo and W mirror specimens, although the quantitative difference is significant for Mo specimens.

Fig. (17). SIMS data for the Zr(46.75)Ti(8.25)Cu(7.5)Ni(10)Be(27.5) alloy samples are given as function of sputtering time of all oxides (a, b, c) and of only beryllium oxide (d) after sputtering by Ar+ ions (a), and after exposure to deuterium plasma ions of energies 1350 eV (b) and 60 eV (c).

The mechanism of the reflectance decrease may qualitatively differ (partially or fully) from the mechanism for Al and Be mirrors. At deuterium interaction with W and Mo mirrors, a transformation occurs of surface transparent oxide layers WO₃ or MoO₃, respectively, into so-called “hydrogen bronze” D_xWO₃ or D_xMoO₃ (e.g., [49, 50]). Unlike oxide properties, these layers with hydrogen as their constituent exhibit electrical and optical properties, which are dependent on the hydrogen concentration. For example, the conductivity of H_xWO₃ may increase by more than seven orders of magnitude as the hydrogen content increases from x=0 up to x=0.4 [51], which undoubtedly leads to a reflectance variation. In other words, the characteristics of the oxide film is transformed from dielectric into nearly metallic.

Experiments on exposure of Cu, Mo and SS mirror samples in the Tore Supra (TS) tokamak [52, 53] have resulted in the discovery of a new effect of mirror interaction with a carbon-containing plasma. The mirrors were arranged close to each other and were open to the plasma during the whole time of their presence in the tokamak chamber, except for a narrow edge around the periphery of the samples, which was closed by the sample holder. This made it possible to perform well calibrated measurements of the layer thickness sputtered under the action of plasma. The greater part of the sample residence time in the tokamak chamber was during wall cleaning procedures by fixed-parameter glow discharges in deuterium or helium. The experimental data on the sputtered layer thickness of the various samples are presented in Table 3.

An important peculiarity of these data is a very low (1.8) experimental value of the ratio of the SS mirror erosion depth to the erosion depth of the Mo mirror. This ratio is nearly a factor of 4 less in comparison to the ratio of SS-to-Mo sputtering coefficients at ion energies typical for glow discharges in TS. During the campaign the anode voltage at operation with helium was +300 V, and with deuterium +400 V. For the corresponding range of ion energy, according to [5], the ratio of sputtering yields Y(SS)/Y(Mo) has to be ~7. An even stronger difference in erosion depth was found for Cu and SS samples: the sputtering coefficients ratio should be Y(Cu)/Y(SS)≡2.5 [5], whereas the obtained ratio of sputtered layer thicknesses equals 12.4.

Fig. (18). Spectral dependence of the difference between initial reflectance and the reflectance after short-term exposure to deuterium plasma ions widely ranging in energy (50 to 1350 eV) for four SC Mo mirrors (a) and two SC W mirrors (b).

Qualitatively similar results were obtained later when exposing mirrors of the same metals in the LHD facility [54]: 1.1 for ratio of SS and Mo erosion depths and 15.5 for the ratio of Cu to SS erosion depths, instead of ~7 and ~2.5, according to data in [5], correspondingly.

It was assumed in [53] that in the presence of a graphite limiter in TS, the main reason for the low sputtering of the SS mirror was a high probability of carbide formation on the mirror surface exposed to a plasma comprising carbon ion admixtures, because all main components of stainless steel (Fe, Cr, Ni) are capable of forming stable carbides. During exposure to the carbon ion-admixed plasma, a thin carbon film, having a lower sputtering coefficient (e.g., see ref [55]), is continuously formed, strongly reducing the sputtering of the metal mirror.

Unlike stainless steel, copper has no stable carbides, and therefore, on exposure to carbon ion-containing plasma the copper mirror shows no film that includes carbon. The copper surface is sputtered by both deuterium (helium) ions and carbon ions. As a result, the surface roughness develops to very high values of 47 nm, as is seen from Table 3.

To verify this hypothesis, a special experiment was made to investigate the behavior of SS and Cu mirror samples in a methane-containing deuterium discharge [56]. After long-term exposure in a deuterium discharge mixed with 3.5% of methane the SS mirror got coated with a rather thick carbon film, which started to flake off after it was exposed some time in air. In contrast, the copper mirror remained clean, but fully lost its reflectance properties because of severe erosion (Fig. 19).

The analysis of data from experiments in which mirrors were exposed inside different fusion devices has shown (see review [62]) that attempt to reduce the influence of the plasma on the mirror through the use of a simple protection (e.g. plane or cylindrical diaphragms) leads to another adverse result: a contaminating film composed of carbon, oxygen and wall materials started to appear on the mirror surface. The reflectance degraded due to contamination of the mirror surface with material eroded from in-vessel components, including the mirror diaphragm, and not due to sputtering. In contrast, the mirrors fully open and exposed to the plasma largely retained their optical properties, provided that the roughness due to sputtering by charge exchange atoms had no time to develop during working discharges and/or by ions during conditioning discharges. Qualitatively similar results were obtained in a comprehensive program to study the behavior of in-vessel mirrors at two versions of JET, with carbon walls [31] and with all metal walls [32].

Fig. (20). Change of normal incidence reflectance at λ=600 nm for W recrystallized (squares) and ITER-grade (solid circles) A.L.M.T. Corp. specimens not irradiated (lines) and irradiated with 20 MeV W+6 ions to 3 dpa (symbols) on the depth of the layer sputtered by 600 eV Ar ions.

In case of a first mirror arrangement as shown in Fig. (1), the FM surface may gradually become contaminated either by the material of walls of the diagnostics channel (if they are clean) or by a deposit composed of wall material. In papers [22, 63] a first attempt has been made to describe qualitatively the mechanism by which the deposition may appear and gradually grow on the FM surface. With known data on material sputtering as basis [5, 64], simple estimations were made for the deposition growth rate versus the angle of incidence of sputtering atoms and their energy, and also, versus the relationship between the channel length and diameter.

Fig. (21a) shows schematically how the mirror at the channel end (“bottom”), facing the plasma confinement volume gets contaminated. In ITER this is the first mirror located at the end of the diagnostic duct (Fig. 1). The sputtering agents here are the charge exchange atoms (in the thermonuclear reactor, these are deuterium and tritium atoms), which have a high energy component sufficient for sputtering the wall material. It is obvious from the scheme (Fig. 21a) that the CXA are incident on the channel wall surface at large angles to the normal, provided that the channel is longer than its diameter. And, as known [64], in this case the flux of sputtered particles substantially increases in comparison with the incidence along the normal, reaching its peak value at approximately 80°. At a sufficiently long channel, only an insignificant part of CXA bombards directly the channel “bottom” surface and sputters away the deposition, so from the estimates in [63] it follows that owing to the wall sputtering, the deposition growth rate considerably exceeds the rate of sputtering by the plasma particles directly arriving the “bottom” of the channel without collision with the channel walls.

The estimations given in [63] show that the relationship between the number of particles sputtered from the channel wall in the direction of the “bottom” surface and the particles deposited onto the “bottom’ can be changed either by making the channel much longer than its diameter, or by considerable shortening it in comparison to the diameter. The first case can be marked as “passive protection”, when particles sputtered from the channel walls are again redeposited on the walls closer to the “bottom” but not on the “bottom” surface itself. The second case can be marked as “active protection”, when sputtering of the “bottom” coating by CXA exceeds the deposition of particles eroded from short channel walls. The case with fully open mirrors is an extreme case when there are no channel walls at all. A high efficiency of CXA in providing suppression of deposit growth on fully open mirrors was shown in mirror experiments in several fusion devices, e.g., Tore Supra [19], TEXTOR [20], LHD [54, 65].

Fig. (21). (a) Scheme of contaminating deposit appearance on the diagnostic channel “bottom” (i.e., on the optical window or the mirror). 1 – peripheral plasma; 2 – CXA flux; 3 – diagnostic channel; 4 – approximate trajectories of sputtered atoms of the diagnostic channel material (or a channel wall deposit); 5 – diagnostic window or mirror; 6 – deposit on the mirror; 7 – approximate trajectories of sputtered mirror deposit atoms. (b) Scheme of how the deposit thickness growth rate can be reduced. The notation 1 to 7 is the same as that for Fig. (22a) 8 – diaphragm from a material with a low sputtering coefficient (e.g., tungsten).

From the above description on how a deposit layer is growing on the “bottom” surface a simple solution follows as to how to substantially reduce the deposit growth rate, or to eliminate the appearance of the deposit at all. For this purpose, it is necessary only to reduce the flux of sputtered atoms from the channel walls reaching the “bottom” surface, to a level lower than the flux of atoms sputtered off from the “bottom” surface by the CXA that arrive from the plasma without collisions with the walls. Qualitatively the same, but a somewhat more complicated structure of the channel is shown Fig. (21b). Here two things are of importance: 1) the diaphragm material must have a low sputtering coefficient (tungsten is best suited for this), and 2) the number of diaphragms and their diameters must be such that there is no direct path of the particles sputtered from the wall to the FM (window) surface.

Recently, a scheme qualitatively similar to that shown in Fig. (21b), has been successfully tested at the Large Helical Device (Toki, Japan) with a simpler shape of fins [54]. It has strongly reduced the mirror contamination, even though stainless steel was used as a material for the diaphragms.

Until recently, carbon-carbon composites have been considered in ITER as material for the divertor target plates (with a total area of 50 m²). Experience of using graphite structural components at numerous fusion devices has shown that additional to charge-exchange atom sputtering, the deposition of carbon-containing film might be another reason of mirror degradation. This is due to the fact that carbon forms volatile molecules with both hydrogen and oxygen, and these molecules may propagate through the vacuum chamber. The threat of carbon deposition is particularly serious for plasma diagnostics mirrors in the divertor region, where the energy of ions and atoms will be low. According to calculations as well as experimental measurements, a pure carbon film of ~10 nm in thickness, can substantially change the spectral reflectance of any metal mirror.

The thickness of carbon-based deposits measured on in-vessel mirrors exposed in the Large Helical Device (LHD), Tore Supra and TRIAM-1M varied in the range 1-50 nm [65]. Therefore, it was necessary to find efficient methods of removing the carbon film directly in-situ in the reactor. One of the methods studied was the usage of a low-temperature hydrogen (deuterium) plasma produced in conditions of electron cyclotron resonance (ECR). It was demonstrated [17] that owing to chemical reactions leading to volatile molecules, this plasma appears highly efficient in removing the carbon film (see Fig. 22) without degradation of the mirror optical properties because of the low ion energy.

Fig. (22). Reflectance recovery (λ=650 nm) of a SS mirror during carbon film (initial thickness ~ 22 nm) removal on exposure time of the mirror to low-energy deuterium plasma. The dashed line shows the reflectance of the clean mirror [17, 70].

A similar procedure with exploitation of a low temperature ECR plasma was successfully used to remove a carbon layer that appeared during exposition of mirrors in the LHD heliotron [66], the TRIAM-1 tokamak [67], in the test stand DSM-2, and similarly, of mirrors exposed in DIII-D and TEXTOR in the toroidal stand TOMAS [68]. A high efficiency of chemical cleaning of hydrocarbon films was demonstrated in ref [69].

Another result that follows from Fig. (22) is that with a decrease in the film thickness, the film erosion rate gradually decreases. This is due to the formation of carbides directly on the SS surface. These are more resistant to the action of hydrogen atoms and ions.

The decision to exclude carbon as a structural in-vessel material in ITER had a considerable effect on the methods that can be utilized for mirror cleaning. This was evident from experiments with full metal walls in ASDEX [71] and JET [32]. Namely, in both devices the mirror specimens in the divertor region and at the outer wall of the main vessel were coated with a deposit that included tungsten, the main material of the divertor tiles. In the case of JET beryllium was one of the main components along with carbon and oxygen [32].

It is not possible to clean mirrors from metal-containing deposits by chemical erosion, as in the case of a hydrocarbon film. At present two methods are under thorough investigation: cleaning by a pulsed laser (e.g., [72, 73] and cleaning by a plasma [74-76]. These techniques have been first tested on their efficiency to remove carbon and hydrocarbon films from the mirrors.

The first successful attempt to clean a mirror exposed in a fusion device (the T-10 tokamak) by a pulsed laser was described in ref [70]. Because of the graphite limiter in use during that experiment, the film was composed of a carbon–hydrocarbon mixture. After removing the samples from the T-10 vessel a full recovery of the initial reflectance of the mirrors was achieved by exposing it to a pulsed excimer laser (λ=308 nm, pulse duration τ=20 ns, repetition frequency f=10 Hz).

The first attempt to clean a Be-containing film was undertaken by the authors of [77]. For cleaning SS and Mo mirrors exposed in JET during 2005-2007 a laser with λ=1064 nm, pulse length τ=120 ns and repetition rate 20 kHz was used. The initial reflectance was not reached for any of the ten mirror specimens tested. Later, single crystal Mo specimens exposed inside the JET vacuum vessel (mainly in the divertor area) during the 2008-2009 campaign were subjected to cleaning with a much shorter laser pulse duration: wavelengths 230, 532 and 1064 nm, pulse length τ=5 ns and repetition rate 20 Hz [78]. For some specimens that had an initially thin contaminating deposit (10-15 nm) a full restoration of the reflectance was achieved.

The experimental results revealed some technical difficulties of laser assisted mirror cleaning. To optimize the process of cleaning, at least five parameters have to be taken into account: laser wavelength λ, pulse duration, laser fluence, pulse repetition rate, and the rate of laser spot scanning along the mirror surface.

In experiments with plasma cleaning of a metal-containing deposit [74-76] aluminium was used as proxy of beryllium to avoid problems with the toxicity of beryllium in laboratory conditions.

In [76] cleaning of an Al (80 nm)+Al₂O₃ (20 nm) film deposited on a single crystal Mo mirror was realized by exposing it in the TOMAS toroidal facility operating with a steady-state ECR-generated He plasma (n_e~3·10¹⁰ cm^-3 and T_e~5 eV) at 2.45 GHz. The plasma ions were accelerated to the mirror by applying a negative voltage -100 V).

The authors of [74, 75] investigated the cleaning of a mirror surface from a deposit composed of Al, Al₂O₃, and W by the use of plasma with as working gas: Ar, Ne or a Ar+D₂ mixture. The most effective in this respect was the Ar plasma produced by a RF capacitively coupled discharge (at frequency 13.56 MHz) with the test mirror utilized as an electrode. In some cases a full cleaning of deposits with initial reflectance restoration was achieved. For the first time a mirror with a size (98 mm in diameter) close to the size of a real first mirror in ITER was successfully cleaned from deposition by a 260 nm thick Al/Al₂O₃ film on its surface [74]. This result is very important when considering the prospects for application of both these techniques (laser and plasma) to remove the contaminating layers from mirror surfaces in ITER environment. With a plasma a large size mirror can be subjected to the cleaning process at once, which results in more homogeneous optical properties along the surface.

The best results with plasma cleaning are expected when the mirror is made of single crystalline metal to avoid the appearance of surface roughness due to sputtering away the contaminating layer by plasma ions.

Experiments with mirrors of different types have added much to the results of paper [79] concerning the efficiency of removing oxide films of different thickness (δ) through the use of low-temperature hydrogen plasma, as is shown in Table 4. All new data (for BeO, Al₂O₃, WO₃ in comparison with the ones of ref [79]) were obtained by applying optical and ellipsometric methods in experiments with metal mirrors (Be, Al, W) and film mirrors (Al, Ag) protected with a dielectric film (Al₂O₃, ZrO₂). The data of Table 4 testify that copper oxide is the most easily removable film, while zirconium oxide film appears the most stable.

To ensure good optical quality of the in-vessel mirrors in ITER during experimental campaigns it is necessary to regularly monitor their optical characteristics. In order to assess whether or not the mirrors can be cleaned one must distinguish between the appearance of a contaminating deposit and the development of surface roughness. In the first case the mirror can be cleaned (e.g. by methods indicated in Sec. 5.2) with restoration of reflectance, but in the second case the measurement results need to be permanently corrected, if mirror cannot be replaced.

One possibility to find the reason why the mirror optical properties have degraded was proposed in ref [45]. The method is based on the analysis of the image of a light source with sharp edges transmitted by the mirror under investigation. In the case the mirror has a deposit of a contaminating film,the reflectance decreases but the edges of the image will continue to be sharp without any diffusive component. In contrast, the appearance of wings in the light source image would indicate the development of surface roughness. The amplitude of the wings increases with increasing roughness of the mirror surface.

The feasibility of this method was demonstrated by its application to mirror specimens coated with carbon film in laboratory conditions and to mirrors coated with contaminated films as a result of exposure of the mirrors in LHD, TRIAM-1M, and Tore Supra. The characteristics of these films are described in refs [65, 66], and [53], respectively.

The effect of long term sputtering by plasma ions on the appearance of diffusive components was studied for mirrors of different materials by the use of the DSM-2 stand (in IPP NSC KIPT). Fig. (23) shows an example of in situ measurements of wings growing under permanent sputtering of a Cu mirror specimen [45]. The thickness of the sputtered layer was estimated by measuring the mass loss after finishing the experiment, under the assumption that the sputter rate does not depend on the sputtering time, which was proven experimentally.

Fig. (23). In situ results obtained when a polycrystalline Cu mirror was gradually sputtered inside the DSM-2 stand. Initially the reflectance raised due to cleaning from oxide and then started to decrease due to roughness development.