Prototype of a radiation hard resistive bolometer for ITER

Transcription

1 INSTITUTE OF PHYSICS PUBLISHING Plasma Phys. Control. Fusion 47 (2005) PLASMA PHYSICS AND CONTROLLED FUSION doi: / /47/12/004 Prototype of a radiation hard resistive bolometer for ITER L Giannone 1, D Queen 2, F Hellman 2 and J C Fuchs 1 1 Max Planck Institut für Plasmaphysik, EURATOM-IPP Association, D Garching, Germany 2 University of California, Berkeley, CA , USA Received 26 July 2005, in final form 8 September 2005 Published 1 November 2005 Online at stacks.iop.org/ppcf/47/2123 Abstract The prototype of a radiation hard resistive bolometer has been produced. This prototype bolometer was installed in ASDEX Upgrade to test its viability and long term stability in a tokamak environment. The prototype bolometer with platinum meanders and absorber on an amorphous silicon nitride substrate and the original standard Kapton bolometer used on ASDEX Upgrade and JET with gold meanders and absorber were calibrated as a function of temperature. The temperature coefficients of the gold and platinum meander resistances are found to have the same value to within 5%. Heat diffusion simulations of the bolometer foils, using the dimensions, specific heat, density and thermal conductivity of the components, were carried out to calculate the cooling time constant and heat capacity of the foils. These calculated values are in agreement with those measured to within 15%. In accordance with these simulations, the prototype bolometer is a factor of 2 more sensitive than the original bolometer and the cooling time constant of the prototype was about a factor of 2 smaller than the original bolometer. The design considerations involved in producing this bolometer foil are discussed and recommendations for future development work are outlined. (Some figures in this article are in colour only in the electronic version) 1. Introduction A schematic diagram of the prototype radiation hard resistive bolometer foil is shown in figure 1. An isolating layer of silicon nitride is grown on a 500 µm thick silicon wafer. This isolating layer of silicon nitride is sandwiched between a metal film absorption layer and a patterned meander track of metal. The temperature rise of the metal film due to the absorbed radiation emitted from the plasma is measured by the change in resistance of the meander. The original bolometer foil used either a Kapton or mica isolation layer with a gold absorber and meander (Mast et al 1991). A 0.2 µm heat conduction layer of gold is deposited onto the /05/ $ IOP Publishing Ltd Printed in the UK 2123

2 2124 L Giannone et al Figure 1. Schematic diagram of a prototype bolometer foil. An isolating layer of low stress silicon nitride (green) with a Pt meander (red) and a Pt absorber (orange) is supported by a silicon frame (light blue). Kapton to improve the thermal contact of the absorber to the bolometer head. The thermal conductivity of gold is two orders of magnitude larger than that of Kapton and therefore this heat conduction layer transports 98% of the heat away from the absorber in the original bolometer foil. A reference and measurement absorber is used to compensate for the increase in meander resistance caused by an increase in bolometer head temperature during a discharge. The silicon frame supports the isolating layer of silicon nitride and provides the thermal contact to the bolometer head. The original bolometer foil has been shown to perform unreliably in an environment with the high neutron fluences expected in ITER (Reichle et al 2001). As the thermal neutron capture cross section of platinum is a factor of 10 smaller than that of gold, it was suggested that the suspected transmutation problem experienced with the gold meander could be slowed down by substitution with platinum. Silicon nitride is currently being investigated as one of a number of ceramics to be used in the ITER environment (González and Hodgson 2003). A resistive bolometer with a platinum absorber and meander on a silicon nitride substrate could then be a suitable radiation hard bolometer detector for installation in ITER. Aspects of resistive metal bolometer foil design and possible alternatives for ITER have also been previously considered (Reichle et al 1996, 1998, Shikama et al 2003). It has also been proposed to measure the temperature rise of the bolometer foil in ITER using thermocouples or optical fibres. For the ITER standard H-mode scenario, it will be necessary to measure plasma radiation with energies up to 25 kev. In figure 2, a plot of the attenuation length of x-rays as a function of energy for Au, Pt and W is shown (Henke et al 1993). A bolometer foil for ITER is therefore estimated to need a Pt absorber thickness of 12 µm to ensure that at least one attenuation length is present to absorb those x-rays with energies up to 25 kev. At the highest energy the absorption efficiency is then 63%. If a higher energy range of absorbed radiation is required then it will be necessary to increase the absorber thickness even further. The actual absorber thickness to use still needs to be established in a more quantitative way. This requires the examination of the various ITER scenarios and their expected radiation profiles and energy spectra to ascertain whether a given absorber thickness produces a measurement of the total radiated power with an unacceptable error.

3 Prototype of a radiation hard resistive bolometer for ITER 2125 Figure 2. Absorption length of x-rays as a function of energy in Au, Pt and W. For the ITER standard H-mode scenario, it will be necessary to measure plasma radiation with energies up to 25 kev. 2. Bolometer foil production The layout of the present bolometer foils was transferred to a printed circuit board layout program. The substitution of the prototype foils into existing bolometer heads with supporting cabling and electronics was facilitated by reproducing the exact dimensions of the present foils. The meander and membrane data was converted by Advanced Reproductions Corporation to produce photomasks of soda lime glass coated with iron oxide. The 30 µm meanders have a ±1 µm resolution on the photomasks. The production of the bolometer foils parallels that of a similar device known as a microcalorimeter. Microcalorimeters have been operated up to 800 K (Denlinger et al 1994) and in magnetic fields up to 8 T (Zink et al 2002). A silicon nitride film with 1.5 micron thickness is grown on a 500 µm thick silicon wafer. The silicon nitride is grown by an LPCVD process (low pressure chemical vapour deposition). Dichlorosilane (SiH 2 Cl 2 ) and ammonia are mixed at 835 C to make a low stress film. This film is non-stoichiometric and Si rich. The platinum meander, with a thickness of 300 or 500 nm, was then sputter deposited onto the silicon nitride film. The silicon was then etched away in the areas of dimensions 4.3 mm 1.8 mm designated for the bolometer foil absorbers. An absorber layer with dimensions of 4 mm 1.5 mm was deposited with thicknesses of either 0.5, 1.0 or 1.5 µm using sputtered deposition through a shadow mask. A total of 30 sample foils were supplied. The laboratory equipment used to produce the prototype bolometer foils is located at the Microfab facility of the University of California, Berkeley. Photographs of the bolometer foil viewed from the meander side are shown in figures 3 and 4. The pair of meanders monitoring the reference absorber is joined to the pair of meanders monitoring the measurement absorber to form a Wheatstone bridge (Giannone et al 2002). The mounting of five foils in a bolometer head has been attempted so far. Three foils were successfully mounted. One failure was due to breaks in the meander structure and the other failure was due to mechanical breakage during mounting. 3. Bolometer measurements Two calibration factors are needed to convert the bolometer signal output of the Wheatstone bridge to an incident power flux. These are the normalized heat capacity, κ, and the cooling

4 2126 L Giannone et al Figure 3. Bolometer foil overview showing one pair of meanders and the interconnections between the reference and measurement foil. The meander width is 30 µm and the thickness is either 300 nm or 500 nm. Figure 4. Close-up view of one meander pair. time constant, τ. The procedure and equipment used for the calibration are detailed elsewhere (see equations A16 and A14 for the definitions of κ and τ, respectively) (Giannone et al 2002). The incident power to the measurement foil is then given by (Mast et al 1991, Giannone et al 2002): with β and g C defined as: P rad = 2 U (R OH +2R C )κ [ g C τ du d dt + u d (1 U 2 )] β 4κ(R OH + R C ) 2 β = 1 (ωcr OH) 2 + (ωcr C ) 2, 1+(ωC(R OH + R C )) 2 (2) g C = 1+(ωC(R OH + R C )) 2 (3) (1)

5 Prototype of a radiation hard resistive bolometer for ITER 2127 and where P rad is the incident power absorbed by the measurement foil, U is the amplitude of applied ac voltage, ω is the frequency of the applied ac voltage, u d is the amplitude of the measured signal, R OH is the meander resistance at the operating temperature of the foil, R C is the cable resistance and C is the cable capacitance. Typical values are g C = for 40 m of cable with C = 2nF,R OH + R C = 1300 and a generator frequency of 19.2 khz. The design of the bolometer foil therefore must take into account the following aspects. The meander resistance should be in the range of This range is a compromise between increasing the resistance to permit the largest amplitude of applied ac voltage for an ohmic heating of the foil that is thermally stable (Vallet and Portafaix 2004) and decreasing the resistance to reduce the RC time constant of the 40 m cables connecting the bolometer amplifier to the bolometer head in the vacuum vessel. The RC time constant determines the value of g C which is related to the signal attenuation along the cable and consequently determines the largest frequency at which the ac voltage can be applied to the Wheatstone bridge of the bolometer foil. The frequency of the applied voltage in turn determines the upper limit of the time resolution of the lock-in amplifier used to demodulate the bolometer signal. Each of the 4 resistances in the Wheatstone bridge formed by the foil must have a ±2% tolerance to operate with the bolometer electronics. The bolometer amplifier operates at gains in the range of 250 to 1000 and to achieve these gains without saturating operation amplifiers it is necessary to compensate the bridge signal in the absence of plasma. The heat capacity of the foil should be as small as possible to ensure the largest temperature rise for a given incident power flux. On the other hand, the smallest possible heat capacity sets an upper limit to the largest amplitude of applied ac voltage possible because of the thermal stability of the foil. The largest amplitude of applied ac voltage ensures the best possible signal to noise ratio has been attained. The heat capacity of a bolometer foil is a product of the volumes of the materials making up the foil and their specific heat. With an applied voltage of 14 V RMS and an ohmic heating of 100 mw per foil, the comparatively small heat capacity of the prototype bolometer means that the temperature rise of the foil due to ohmic heating is 173 C. In contrast, the temperature rise of the gold and Kapton bolometer due to ohmic heating is 75 C. The choice of cooling time constant is a compromise between better time resolution with lower values and improved sensitivity with larger values. Despite a cooling time constant in the vicinity of 100 ms, a time resolution of 3 ms is possible with appropriate signal processing. In the W7-AS HDH mode, jumps in the radiated power due to impurity injection over such a time scale could be observed (Giannone et al 2003) (see figure 2). It is expected that the design criteria for the 10 ms time resolution of the radiated power measurement for the main plasma and divertor of ITER (Costley and ITER National Team, Naka 2005) can be easily satisfied using the proposed bolometer with 12 µm Pt absorber. Physically, the cooling time constant of the bolometer foil depends on the thermal conductivity and thickness of the materials connecting the absorber to the supporting frame and the distance from the absorber to the supporting frame. In the prototype bolometer, the choice of SiN membrane thickness of 1.5 µm was governed by the availability of this thickness as a standard process at the Microfab facility. The distance from the absorber to the silicon frame of 0.15 mm was considered as a safe choice for a cooling time constant that ensured that the operating temperature of the bolometer foil was not too high for the applied ac voltage of 20 V amplitude. The foil dimensions were chosen so that the prototype foil could be used as a substitute for the original bolometer foil. An existing bolometer head using a prototype foil was installed in ASDEX Upgrade. In figure 5, the time evolution of the bolometer signals in a discharge at 2.5 T with a plasma current of 800 ka and 5 MW neutral beam heating is plotted. The power flux

6 2128 L Giannone et al Figure 5. Time evolution of signals from the prototype bolometer in a discharge at 2.5 T with I p = 800 ka and 5 MW neutral beam heating. Figure 6. Temperature dependence of the platinum meander resistance of the prototype bolometer. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. incident onto each absorber viewed along the four tangential lines of sight can be calculated from these signals. Five months of continuous operation without failure indicate that the prototype bolometer foil is robust enough for installation in a tokamak. 4. Meander resistance In figure 6, the temperature dependence of the platinum meander resistance of the prototype bolometer is plotted. In figure 7, the temperature dependence of the gold meander resistance of the original bolometer is plotted. The line of best fit for the resistance, R, in terms of the temperature, T,isgivenby R = R 0 (1+α(T 20)), (4) where R 0 is the resistance at T = 20 C and α is the temperature coefficient of resistivity (TCR). The TCR, of the platinum meander ( ) is roughly 5% greater than that of the

7 Prototype of a radiation hard resistive bolometer for ITER 2129 Figure 7. Temperature dependence of the gold meander resistance of the original bolometer. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. Table 1. A comparison of the neutron capture cross section, resistivity at T = 20 C and TCR of candidate metals for the bolometer meander. Metal Neutron capture cross section (barns) ρ (µ cm) TCR (K 1 ) Au Pt Ni W gold meander ( ) but both are distinctly lower than the theoretically possible value (see table 1). It has been suggested that one possible cause of this smaller value is the Matthiessen law (Maissel and Glang 1990). As the resistivity of the deposited material is the sum of the resistivity due to the bulk materials and the resistivity due to the defects, it is expected that only the contribution due to the bulk material has the proper temperature dependence and that the TCR of the thinner resistor is dominated by the contribution due to the defects. Indeed, increasing the Pt meander thickness to a nominal value of 500 nm increased the TCR to Also, a 10 nm thick film of Ni had a TCR of , about a third of the bulk value (Brunetti and Monticone 1993). 5. Membrane without absorber This calibration procedure was applied to a silicon nitride membrane of nominal thickness 1.5 µm without platinum absorber and the calibration factors were measured as a function of temperature. In figure 8, the temperature dependence of κ for the silicon nitride membrane without absorber is plotted. In figure 9, the temperature dependence of τ for the silicon nitride membrane without absorber is plotted. In these plots, the values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown.

8 2130 L Giannone et al ( C) Figure 8. Temperature dependence of the normalized heat capacity, κ, of the silicon nitride membrane without absorber. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. ( C) Figure 9. Temperature dependence of the cooling time constant, τ, of the silicon nitride membrane without absorber. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. These results can be used to compare with simulations using the known values of heat capacity (2.1 MJ m 3 K 1 ) and thermal conductivity (3.2 W m 1 K 1 ) of a silicon nitride membrane (Mastrangelo et al 1990, Zink and Hellman 2004). At 50 C, a value of 24.4 µjk 1 is calculated for the 1.5 µm thick membrane with a Pt layer of 0.15 µm, which is half the thickness of the meander. This is in fair agreement with the experimental value of 26.8 µjk 1. The simulated value of τ = 97 ms is in good agreement with the experimentally measured value of 93 ms. The increase in κ and τ with temperature can be understood in terms of an increasing specific heat of silicon nitride at temperatures above 20 C (Zink and Hellman 2004). Even though the thermal conductivity of silicon nitride also increases with temperature, the measured increase in τ with temperature indicates that the increase in specific heat with temperature dominates.

9 Prototype of a radiation hard resistive bolometer for ITER 2131 Figure 10. Temperature dependence of the normalized heat capacity, κ, of the prototype bolometer. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. Figure 11. Temperature dependence of the normalized heat capacity, κ, of the original bolometer with Kapton isolator. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. 6. Calibration In figure 10, the temperature dependence of the normalized heat capacity of the silicon nitride membrane, with platinum meander and 1.5 µm absorber, of the prototype bolometer is plotted. In figure 11, the temperature dependence of the normalized heat capacity of the Kapton isolator, with gold meander and 4 µm absorber, of the original bolometer is plotted. The normalized heat capacity of the prototype bolometer is a factor of 2 smaller. For the same power flux on the bolometer, the output signal is then a factor of two larger. In figure 12, the temperature dependence of the cooling time constant of the silicon nitride membrane, with platinum meander and 1.5 µm absorber, of the prototype bolometer is plotted. In figure 13, the temperature dependence of the cooling time constant of the Kapton isolator, with gold meander and 4 µm absorber, of the original bolometer is plotted. The cooling time

10 2132 L Giannone et al Figure 12. Temperature dependence of the cooling time constant, τ, of the prototype bolometer. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. Figure 13. Temperature dependence of the cooling time constant, τ, of the original bolometer with Kapton isolator. The values and lines of best fit for the measurement (dark green) and reference foils (light green) are shown. constant of the prototype bolometer is a factor of 2 smaller. This cooling constant is smaller than desired. This first set of photomasks provided only 0.15 mm rather than the full 0.5 mm distance between the absorber edge and silicon frame. The next batch of prototype bolometers will use photomasks modified to increase the cooling time constant. The origin of the temperature dependences of the cooling time constant and normalized heat capacity of the prototype bolometer foil is the increase in the specific heat of the low stress silicon nitride membrane with temperature (Revaz et al 2003). This effect dominates the increase in thermal conductivity with temperature of the silicon nitride which would tend to lead to a decrease in τ with increasing temperature. In the original bolometer foil the same observed trend of the cooling time constant and normalized heat capacity with temperature was also previously measured (Schubert 1999). This trend can similarly be explained by the

11 Prototype of a radiation hard resistive bolometer for ITER 2133 increase in heat capacity of the components with temperature (Buch 1999). Additionally, the thermal conductivity of the gold heat conducting layer decreases with temperature (Buch 1999), leading to a further contribution to the increase in the cooling time constant with temperature. A comparison of the quantity 1/τ τ/ T, with a value of K 1 for the prototype bolometer foil and K 1 for the original bolometer foil with Kapton isolator, indicates that the cooling time constant of the prototype bolometer foil has a reduced temperature dependence. This is due to the fact that the thermal conductivity of SiN increases with temperature tending to reduce the cooling time constant and this compensates in part the increase in specific heat with temperature tending to increase the cooling time constant. A comparison of the quantity 1/κ κ/ T, with a value of K 1 for the prototype bolometer foil and K 1 for the original bolometer foil with Kapton isolator, indicates that the normalized heat capacity of the prototype bolometer foil has an increased temperature dependence. This temperature dependence of the calibration constants must be accounted for when operating a bolometer camera to obtain reliable measurements for tomographic inversion of the measured line integrals of power flux. A study of the temperature evolution of the bolometer cameras in ITER has also demonstrated that the calibration constants are a function of temperature (Vallet and Portafaix 2004). In a long pulse tokamak like ITER, it will be necessary to match the meander resistance, cooling time constant and heat capacity of the measurement and reference foil as closely as possible, so that in the event of a long term temperature rise of the bolometer head, the consequent drift in the signal output is minimized. This aspect is especially relevant to a difference in the nuclear heating of the Pt measurement and reference absorber arising from possible differences in the volume of the absorber. Assuming the worst case nuclear heating of 10 MW m 3 given for the tungsten coating in the divertor (Iida et al 2004) anda1% difference in volume of a 12 µm thick absorber, this would then be equivalent to an incident power flux of 1.2 W m 2. This is typically the noise level of a bolometer measurement with a signal level of 100 W m 2. The nuclear heating value used for tungsten on the divertor plate is an overestimate of the value expected for the bolometer foils located further from the plasma. A thermal analysis of the operation of the bolometer cameras in ITER including the contributions due to the nuclear heating of the materials used to mount the cameras is available (Vallet and Portafaix 2004). The operation of bolometer cameras viewing a single absorber foil with an infrared camera (Wurden and Peterson 1999) is seriously compromised in ITER as the ability to compensate for the nuclear heating of the absorber is absent. It is to be expected that during the lifetime of the bolometer foil for the measurement and reference absorber the values of meander resistance, cooling time constant and heat capacity will change with the accumulated neutron dose. An in situ test is required to test if the neutron dose in ITER causes these values for the measurement and reference absorber to drift too far apart. In such a case, this would limit the useful lifetime of the bolometer foil. In tables 2 and 3, the values of the meander resistance, R, normalized heat capacity, κ, and cooling time constant, τ, for the measurement (subscript M) and reference (subscript R) foil of the prototype and original bolometer, respectively, are given. The difference between the measurement and reference value is then normalized to the mean and shown in table 4. Presently, the normalized difference in values of meander resistance, cooling time constant and normalized heat capacity of the measurement and reference foil is in each case larger in the prototype bolometer. The ability to laser trim the meander resistance has been investigated. Preliminary tests show that a commercially available laser trimming system (LS Laser Systems, Munich) is capable of changing the values of the meander resistance. Optimization of the laser properties will be needed to ensure minimal damage of the silicon nitride membrane. The purchase of

12 2134 L Giannone et al Table 2. The calibration constants at T = 50 C for the measurement (M) and reference (R) foils of a prototype silicon nitride and platinum bolometer. Channel R M ( ) R R ( ) κ M (A 2 ) κ R (A 2 ) τ M (ms) τ R (ms) E E E E E E E E Table 3. The calibration constants at T = 50 C for the measurement (M) and reference (R) foils of an original Kapton and gold bolometer. Channel R M ( ) R R ( ) κ M (A 2 ) κ R (A 2 ) τ M (ms) τ R (ms) E E E E E E E E Table 4. The normalized difference of the calibration constants at T = 50 C of the measurement and reference foils of a prototype silicon nitride and platinum bolometer (left) and an original Kapton isolator and gold bolometer (right). Channel δ R δ κ δ τ δ R δ κ δ τ such a laser trimming system is required for further development of this technique on a future bolometer prototype. 7. Heat diffusion simulations Numerical simulations of two dimensional (2D) heat flow in thin films is a common tool for extracting the heat capacity and thermal conductivity from a thin film silicon nitride membrane. The dimension of the foil in the vertical direction is much smaller than those in the horizontal plane and consequently the diffusion time in this direction is much faster than that in the horizontal direction. Therefore, the temperature can be assumed to be constant in the vertical direction. The 2D heat diffusion equation is (Revaz et al 2003): c 2D (x, y) T(x,y,t) [ ( k 2D (x, y) T(x,y,t) ) + ( k 2D (x, y) T(x,y,t) )] t x x y y = P 2D (x,y,t) (5) with k 2D = kt, c 2D = ρtc and P 2D = P/S and where k[w K 1 m 1 ] is the thermal conductivity, C[J K 1 kg 1 ] is the specific heat per unit mass, ρ[kg m 3 ] is the density, t[m] is the film thickness and P/S[W m 2 ] is the power flux to the absorber. It is assumed that k 2D or c 2D is the sum of the values for each component material when they are joined together. The boundary condition on the edge of the membrane is T = 0. This partial differential equation

13 Prototype of a radiation hard resistive bolometer for ITER 2135 Temperature (K) µm Au + Mica 4µm Au + Kapton 1.5µm Pt + SiN 4 µm Pt + SiN 12µm Pt + SiN Time (s) Figure 14. The time evolution of the mean temperature of the absorber when switching on a 40Wm 2 source at t = 0 s is plotted for various bolometer foils. These are the original Kapton and gold bolometer (green) and the prototype silicon nitride and platinum bolometer (purple) with reduced distance between the silicon frame and absorber. The time evolution of the mean temperature of the absorber for bolometers on a low stress silicon nitride membrane with a 4 µm (red) and 12 µm (orange) Pt absorber assuming a 0.5 mm separation between absorber and silicon frame and a bolometer with a mica isolator and 4 µm gold absorber (blue) is also shown. The time constants, heat capacities and steady state temperatures reached are summarized in table 5. is solved on a regular grid using the forward-time centred-space finite difference equation (Hoffman 1993). Symmetry permits the equations to be solved in one quadrant of the foil. In figure 14, the results of the simulations of various bolometer foils are shown. Using the specific heat, density and thermal conductivity of Kapton and gold, the time evolution of the mean temperature of the absorber of the original bolometer foil (green) was calculated. Using the values of specific heat, density and thermal conductivity of low stress silicon nitride and platinum, the time evolution of the mean temperature of the absorber of the bolometer foil (purple) with 0.15 mm between the absorber edge and silicon frame was calculated. In both cases it is assumed that a source of 40 W m 2 is switched on at t = 0 s. The time evolution of the mean temperature of the absorber for bolometers on a low stress silicon nitride membrane with a 4 µm (red) and 12 µm (orange) Pt absorber assuming a 0.5 mm separation between absorber and silicon frame and a bolometer with a mica isolator and 4 µm gold absorber (blue) is also shown. The bolometer foil with mica isolator and 8 µm gold absorber planned for the JET enhanced performance phase (McCormick et al 2005) has a heat capacity only 25% larger than that of a 4 µm absorber as the greater part of the heat capacity is in the mica isolator. The time constants, heat capacities and steady state temperatures reached in each case are summarized in table 5. The modelled and experimental values (see tables 2 and 3) of the cooling time constant are in good agreement for both bolometers. The experimental values of heat capacity at 50 C are 40 µjk 1 for the prototype and 160 µjk 1 for the original bolometer. The cooling time constant is 60 ms for the prototype bolometer and 110 ms for the original

14 2136 L Giannone et al Table 5. A comparison of the cooling time constants, heat capacities and saturated temperatures reached for the simulated bolometer foils. The measured cooling time constants from tables 2 and 3 are in good agreement with these values calculated in the simulations. The experimental values of heat capacity at 50 C are 40 µjk 1 for the prototype and 160 µjk 1 for the original bolometer. Hence, the nominal values of the dimensions, specific heat, density and thermal conductivity of the components are sufficient to reproduce the experimental values of heat capacity to within 15%. Absorber Isolator τ (ms) Heat capacity (µjk 1 ) Temperature (K) 4 µm Au 7.5 µm Kapton µm Pt 1.5 µm SiN µm Pt 1.5 µm SiN µm Pt 1.5 µm SiN µm Au 20µm Mica bolometer. As the 2D heat diffusion simulations use only the nominal values of the component dimensions and the literature values for their specific heat, density and thermal conductivity (Mastrangelo et al 1990, Revaz et al 2003, Weast 1970, Data Sheet 2005) only a qualitative agreement could be expected with the measured heat capacity and cooling time constant. However, the measured and calculated values of heat capacity and cooling time constant are in agreement to within 15%. This confirms that the performance of future designs can be predicted with adequate precision. Increasing the distance of the absorber to the silicon frame to 0.5 mm will increase the cooling time constant to 227 ms and increase the sensitivity of foils with a 4 µm Pt absorber by a factor of 3.6 compared with the currently used Kapton and gold bolometer foil with a 4 µm absorber. It is projected that the ITER bolometer with a 12 µm thick platinum absorber would have a cooling time constant of 225 ms and a factor of 1.5 improvement in sensitivity compared with the currently used Kapton and gold bolometer foil. Corrections for thermal radiation from the bolometer foil surface can be carried out to ascertain whether the power emitted needs to be accounted for in the calibration procedure. A first order estimate of the radiation correction, P SB, is determined by adding terms for the emission of radiation from the heated area of the bolometer absorber and absorption of radiation from the environment (kept at T 0 by a radiation shield) (Zink et al 2005): P SB = Aɛσ ((T 0 + T ) 4 T0 4 ), (6) where T is the temperature rise of the absorber, A is the emitting area, ε is the emissivity of the surface and σ = Wm 2 K 4 is the Stefan Boltzmann constant. For the simulations with foil temperature equal to environment temperature, then for T 0 = 300 K, ε = 0.10 and A = m 2 the radiation correction term to first order is P SB = 3.6 µwk 1 T. The area is assumed to be simply equivalent to that of the absorber as the emissivity of silicon nitride is greater than that of the metal absorber. A uniform temperature over the absorber area is assumed for this estimate. It is concluded that this correction term is 8% of the prototype bolometer foil nominal heat capacity of 46 µwk 1 from calculations (see table 5). Neglecting the radiation correction term leads to a systematic overestimate of the heat capacity as the measured temperature rise will be smaller than expected because part of the ohmic heating power is being radiated. The radiation correction term will be of reduced significance for bolometers with thicker absorbers. At T 0 = 400 K, the radiation correction term to first order is P SB = 8.5 µwk 1 T. At the higher temperature the overestimate of the heat capacity of the bolometer is greater and this means up to 4.9 µwk 1 of the increase in the measured heat capacity could be due to the radiation correction term. Over the temperature range of 300 K to 400 K the heat capacity of

15 Prototype of a radiation hard resistive bolometer for ITER 2137 Time (s) Time (s) Figure 15. The time evolution of the bridge voltage and absorbed power of a mica bolometer illuminated with a HeNe laser. The raw data of the measured bolometer bridge voltage are (a) without the FIR filter and (b) with the FIR filter of the AD7730. the prototype bolometer foil increases from 40 µwk 1 to 53 µwk 1 and the standard Kapton isolator with gold absorber and meander increases from 150 µwk 1 to 190 µwk 1. These estimates indicate that the radiation correction term is not sufficient to explain the measured temperature dependence of the bolometer heat capacity. 8. Bolometer amplifier development A bolometer amplifier with a 2 khz square wave and an amplitude of 4.1 V applied to the bolometer has been developed. Using the AD7730 (Analog Devices), a 19 inch rack with 32 channels can be interfaced through DIO connections to a PXI 7833R (National Instruments). It is envisaged that 64 channels can be connected to one PXI module. In comparison with the ac system previously described (Giannone et al 2002), the offset compensation is greatly simplified. The AD7730 has the ability to correct the dc bridge offset. Shown in figure 15 is the bridge voltage measured when illuminating a mica bolometer in the laboratory with a HeNe laser of 2 mw nominal power output. Using the bolometer equation of equation 1, the bridge voltage can be transformed into the power absorbed by the bolometer foil. The absorbed power measured is consistent with the reflection expected from the gold foil at the laser wavelength of nm and the expected transmission through the glass window of the vacuum vessel (Schubert 1999). In figure 15, a comparison is made between the two operating modes of the AD7730 for the measured bridge voltage and absorbed power. The measurements of the bridge voltage are (a) without the finite impulse response, FIR, digital filter or (b) with the FIR filter of the AD7730. The sample rate and chop frequency is 2 khz without the FIR filter and 1 khz with the FIR filter. The bandwidth is reduced from 500 Hz without the FIR filter to 40 Hz with the FIR filter. When bypassing the FIR filter on the chip, the measured RMS noise voltage is 3.0 µv and with the FIR filter the RMS noise is 1.0 µv. The expected RMS value of the thermal noise, V n = (4kTRB) = 0.12 µv, for an R = 1600 load at T = 300 K and bandwidth, B, of 500 Hz. From the data sheet for the AD7730, 0.7 µv is the expected RMS value at a sample rate of 2 khz. Both these values of RMS noise are better than the value of 20 µv measured on the channels in figure 5 using the current ASDEX Upgrade bolometer amplifiers. It should

16 2138 L Giannone et al Time (s) Time (s) Figure 16. The time evolution of the bridge voltage and absorbed power of a mica bolometer illuminated with a HeNe laser on an expanded time scale. The derivative of the bridge voltage is performed with (a) a differentiating Bessel filter and (b) a differentiating Lanczos or Savitsky Golay filter. Figure 17. A comparison of the frequency response of a differentiating Bessel filter and a differentiating Lanczos or Savitsky Golay filter. be noted that the signal levels at the bridge will be 3.5 times smaller with these bolometer amplifiers as the applied voltage is also a factor 3.5 smaller than the current ASDEX Upgrade bolometer amplifiers. Using the FIR filter reduces the bandwidth to 40 Hz and therefore reduces the expected thermal noise to 0.03 µv. From the data sheet for the AD7730, 0.2 µv are specified at a sample rate of 1 khz when employing the FIR filter. In the laboratory, the noise level measured is then a factor of four to five above the specifications. It remains to be proved that these values can be obtained in the experimental hall in the vicinity of ASDEX Upgrade. In figure 16 a comparison is made between two types of differentiating digital filter to calculate the absorbed power using the data taken without using the FIR filter of the AD7730. These filters are (a) a second order low pass Bessel filter and a differentiating Bessel filter at 30 Hz and (b) a differentiating Savitsky Golay (or Lanczos) filter (Lanczos 1988, Hamming 1989). The frequency response of each filter is shown in figure 17.

17 Prototype of a radiation hard resistive bolometer for ITER 2139 Figure 18. A comparison of the coefficients of a differentiating Bessel filter and a differentiating Lanczos or Savitsky Golay filter for convolution with the measured bridge voltage. When using the FIR filter, the time delay to reach the mean value of power is 15 ms rather than the 10 ms seen in figure 16. This increase in time delay is a consequence of the constant number of terms used in the differentiating filter. Halving the number of data points involved in calculating the derivative would increase the noise level. The advantages of using the Savitsky Golay differentiating digital filter are twofold. Firstly, as shown in figure 18, the length of the convolution function is considerably shorter. Secondly, the better frequency response of the differentiating Savitsky Golay filter more successfully damps out the higher frequency components. The transformation of the measured bridge voltage to absorbed power has been implemented on the PXI 7833R s FPGA. This allows the line integrals of radiated power for each channel of the bolometer camera to be communicated in real time to the machine control system. The conversion of the bolometer bridge signal to absorbed power with analog feedback electronics has also been demonstrated (Schivell et al 1982). The second input to the AD7730 is used to measure the calibration constants of the bolometer. The procedure for obtaining these values has been discussed in detail elsewhere (Giannone et al 2002). The variation of the calibration constants obtained for a single bolometer channel in each of the 32 bolometer amplifiers is summarized in table 6. Repeating the measurements for a single channel shows that the cooling time constant and normalized heat capacity measurements have a standard deviation comparable to the value given when comparing different amplifiers. The ultimate goal is to measure the calibration constants with a reproducibility of better than 1%. The next round of development will aim to improve the reproducibility. A sensitivity of 8 V W 1 will be reached by using the 4 µm Pt foil and increasing the square wave amplitude to 9 V. This would be equal to the sensitivity achieved by the bolometer foil with mica isolator and 8 µm gold absorber and ac electronics with 40 V pp applied voltage installed for the JET enhanced performance phase (McCormick et al 2005). The reported noise level for the ac electronics of 20 mv pp at a gain of 5000 is equivalent to 0.66 µv RMS at the bridge and therefore slightly smaller than the 1.0 µv RMS at the bridge with the AD7730 employing the FIR filter. However, the power dissipation will be a factor of 10 smaller and the temperature rise of the foil due to ohmic heating will be a factor of 2.5 smaller in the prototype bolometer foils and electronics with 9 V square wave amplitude. The claim that the sensitivity and cooling time constant of the bolometer foils in JET are independent of temperature is not in agreement with measurements made on a bolometer foil with mica isolator and 4 µm gold absorber. Additionally, the ac electronics delivers only the raw bolometer signal that still

18 2140 L Giannone et al Table 6. A comparison of the mean and standard deviation, σ, of the meander resistance, R, normalized heat capacity, κ, and cooling time constant, τ, for the measurement and reference bolometer using the same bolometer channel in each of the 32 bolometer amplifiers. The standard deviation of measurements of a single channel, σ S, are comparable for κ and τ. Measurement σ σ S Reference σ σ S R( ) κ (10 3 A 2 ) τ (ms) must be processed into incident power flux. If the prototype bolometer foil were operated at the same temperature as the JET bolometer foil by increasing the square wave amplitude to 14 V, then the improvement in sensitivity by a factor of 1.5 would compensate the factor of 1.5 higher noise level of the AD7730 and the equivalent detection limit of 1 µwcm 2 as the JET system could be attained. 9. Discussion Further improvements to the bolometer foil design are possible to make it more suitable for operation in ITER. As shown in table 1, the neutron capture cross section of Pt is approximately a factor of 10 smaller than that of Au (Firestone 2005). A further reduction by a factor of 2 would be possible by using Ni as the material for the meander. In addition, the TCR of Ni is higher than that of Pt or Au and this would further increase the bolometer sensitivity. Such an Ni thin film resistor has been tested up to 400 K (Brunetti and Monticone 1993). Also listed in table 1 are the resistivity at T = 20 C and the TCR of these metals (Buch 1999). The deposition of Ni meanders on a SiN membrane is not yet a standard process. However, the operating temperature of the bolometer foil must be less than 550 K, as above this temperature a ferromagnetic transition in nickel results in a decrease of the TCR (Zao and Tsai 1986). Thermal analysis of the operation of the bolometer cameras in ITER, including the contributions due to the nuclear heating of the materials used to mount the cameras, shows that an uncooled camera may have a temperature rise of 350 K over the duration of a 1600 s pulse (Vallet and Portafaix 2004). It is known that He generation from Ni at high neutron doses above 10 dpa can accumulate at grain boundaries and make the steel alloy brittle at temperatures above 500 C (Garner et al 1998, Budylkin et al 2004). A test comparing Ni and Pt meanders in an ITER relevant neutron dose is also needed to clarify which candidate material has the longer lifetime as thin film nickel exposed to fast neutron irradiation has been observed to suffer from localized lattice transformations due to thermal spikes (Teodorescu and Giodeanu 1960). The melting point of Pt (1772 K) is higher than that of Ni (1453 K). The phenomenon of radiation induced conductivity, RIC, has been investigated for Si 3 N 4 (Snead et al 1994) and Al 2 O 3 (Snead et al 1994, Shikama et al 1998). At a dose rate of 10 4 Gs 1, a value of conductivity of m 1 is given. However, only a dose rate of 100Gs 1 is expected at the location of the bolometers in ITER (Costley et al 2001). Assuming that the low stress silicon nitride possesses similar RIC characteristics to those ceramics already measured, then at this dose rate it is estimated that for a 1.5 µm thick membrane, a resistance of 1.2 M will still be achieved between the meander and absorber. An investigation of the RIC for low stress silicon nitride membrane used in the prototype bolometer foil has not yet been carried out. Presently, spring loaded pins are used to make the electrical contact from the cables to the bolometer foil. The loss of spring tension was suspected as being the reason for the failure

19 Prototype of a radiation hard resistive bolometer for ITER 2141 of contacts to the foil in earlier irradiation experiments (Reichle et al 2001). A more robust method of contacting, capable of withstanding the temperature cycles and neutron fluences of ITER, needs to be developed. A method of platinum bonding these contacts, similar to gold bonding used in recent generations of JET divertor bolometers, has been proposed to improve the radiation hardness of the bolometers to be installed in ITER. An in situ test of the prototype bolometer foil in a bolometer head in the presence of an ITER relevant neutron fluence on the SCK-CEN irradiation facility (Gusarov et al 2005) is planned. The radiation hardness of a bolometer head with a low stress silicon nitride isolator and Pt meanders must be demonstrated before use in ITER. Furthermore, there is evidence to suggest that even at low neutron doses the thermal conductivity of ceramics is altered (Snead et al 2004). The thermal conductivity of Si 3 N 4 is reduced by a factor of 2 by a neutron dose of 0.1 dpa. In a 1000 s pulse of ITER, a total dose of 10 5 dpa is expected in ceramics at the position of the bolometer foil (Costley et al 2001). Measurement of the calibration constants during a single discharge would be necessary only if the dose were sufficient to significantly change them in that discharge. In an ITER disruption, a large fraction of the total energy content will be radiated in a short time. It is necessary to assess whether this large power flux is sufficient to cause an intolerable temperature rise of the absorber and damage the bolometer foil. In the limit of time scales much shorter than the cooling time constant the relationship between absorbed power, P rad, and temperature rise of the foil, T, in a time interval, t, isgivenby P rad = C T, (7) t where C is the heat capacity of the bolometer foil. Nuclear heating of the Pt foil at 10 MW m 3 (Vallet and Portafaix 2004) can be neglected in this estimate. If a temperature rise of 400 K is assumed to be acceptable then, for the ITER bolometer foil considered with an absorber area of 4.94 mm 2, the critical flux over the assumed 1 ms duration of a gas pulse mitigated disruption is estimated to be 20 MW m 2. This power flux is to be compared with 0.25 MW m 2 estimated to be the value at the first wall in such a scenario (Vayakis 2005). The critical lines of sight will be those with long path lengths through the plasma edge where the highest radiated power densities would be expected. 10. Conclusions A prototype radiation hard resistive bolometer has been built and has now been in operation for five months in a tokamak environment. A 1.5 µm isolating layer of low stress SiN provides the basis for the 1.5 µm Pt absorber of plasma radiation and the Pt meanders that measure the temperature rise of the absorber. The calibration of bolometer foils as a function of temperature reveals that the normalized heat capacity, κ, and cooling time constant, τ, increase with temperature. This feature can be explained in terms of the increase in specific heat of the component materials with temperature. This effect dominates the increase in thermal conductivity with temperature which would tend to lead to a decrease in τ with increasing temperature. This temperature dependence of the calibration constants must be accounted for when operating a bolometer camera to obtain reliable measurements for tomographic inversion of the measured line integrals of power flux. A comparison of these measured properties with those calculated by 2D heat diffusion simulations confirms the ability to predict the performance of future designs. The radiation correction term will be of reduced significance for bolometers with thicker absorbers.

20 2142 L Giannone et al Further improvements to the prototype bolometer foil will be the aim of the next round of foil production. The prototype foil will have an absorber thickness close to that required for ITER operation. A prototype bolometer foil with an Ni meander is proposed, since this would further reduce the thermal neutron capture cross section and increase the bolometer foil life time in ITER. Also, the bolometer sensitivity with an Ni meander would be increased, as the TCR of Ni is larger than that of Pt. However, the deposition of Ni meanders on a SiN membrane is not yet a standard process and above 550 K the TCR is reduced. Increasing the distance of the absorber to the silicon frame to 0.5 mm will increase the cooling time constant to 227 ms and increase the sensitivity of foils with a 4 µm Pt absorber by a factor of 3.6 compared with the currently used Kapton and gold bolometer foil with a 4 µm absorber. To produce bolometer foils with the lowest possible thermal drift necessary for ITER long pulse operation, it will be necessary to better match the heat capacity, cooling time constant and meander resistances of the measurement and reference foil. Laser trimming of the meander resistances has been demonstrated to be a feasible method of achieving the desired resistance tolerance. The purchase of such a laser trimming system is required to further develop this technique on a future bolometer prototype. The actual absorber thickness to use on ITER still needs to be established. This requires the examination of the various ITER scenarios and their expected radiation profiles and energy spectra to ascertain whether a given absorber thickness produces an unacceptable measurement of the total radiated power. An in situ test of the prototype bolometer foil in a bolometer head in the presence of an ITER relevant neutron fluence to demonstrate its radiation hardness is planned. The spring loaded electrical contacts used to mount the foils in current bolometer heads limit the radiation hardness of the present construction. Development work is needed to establish a radiation hard method of contacting the cable to the foil. For an ITER disruption, it is necessary to assess whether the large power flux produced on a short time scale will cause an intolerable temperature rise of the absorber and damage the bolometer foil. Acknowledgments This work was financially supported by EFDA contract Useful discussions with A Costley, L C Ingesson and JCVallet have led to improvements in many aspects of the results presented. The laboratory infrastructure and mounting of the foils essential for this work was expertly provided by H Wolf. The development of the bolometer amplifiers is the concientious work of A Bauer, H Eixenberger and G Lexa. References Brunetti L and Monticone E 1993 Meas. Sci. Technol Buch A 1999 Pure Metal Properties (Ohio: ASM International) Budylkin N I et al 2004 J. Nucl. Mater Costley A E, Campbell D J, Kasai S, Young K E and Zaveriaev V 2001 Fusion Eng. Design Costley A E and ITER National Team 2005 ITER measurement requirements Protected stuff/eighth ITPA Meeting/Costley.requirements.pdf. Data sheet 2005 Physical and thermal properties of Kapton Denlinger D B et al 1994 Rev. Sci. Instrum Firestone R W 2005 PGAA Database Garner F A, Oliver B M and Greenwood L R 1998 J. Nucl. Mater Giannone L et al 2003 Plasma Phys. Control. Fusion