BEHAVIOUR OF INTRINSIC CARBON AND LASER BLOW-OFF INJECTED NICKEL IN TORE SUPRA DURING ERGODIC DIVERTOR ACTIVATION

Transcription

1 FR DRFCCAD EUR-CEA-FC-59 BEHAVIOUR OF INTRINSIC CARBON AND LASER BLOW-OFF INJECTED NICKEL IN TORE SUPRA DURING ERGODIC DIVERTOR ACTIVATION M. Mattioli, C. de Michelis, P. Monier-Garbet Aout 994 ASSOCIATION EURATOM-C.E.A. DEPARTEMENT DE RECHERCHES SUR LA FUSION CONTROLEE C.E.N.CADARACHE 308 SAINT PAUL LEZ DURANCE CEDEX

2 BEHAVIOUR OF INTRINSIC CARBON AND LASER BLOW-OFF INJECTED NICKEL IN TORE SUPRA DURING ERGODIC DIVERTOR ACTIVATION M. MATTIOLI, C. DE MICHELIS and P. MONIER-GARBET Association Euratom-CEA sur la Fusion Departement de Recherches sur la Fusion Controlee CEN-Cadarache, F-308 St Paul-lez-Durance ABSTRACT The behaviour of intrinsic carbon and laser blow-off injected nickel has been studied in Tore Supra during ergodic divertor (ED) activation. C VI and C V line intensiy ratio simulations require in the plasma ergodic edge both increased diffusion and a large (in the lo^-lo 7 m-3 range) neutral hydrogen isotope density. A peripheral barrier has to be introduced to simulate Ni injections. If it is taken purely diffusive, satisfactory C line ratios can be recovered, but the simulated injected impurity confinement time ip is too short. For a satisfactory simulation it is necessary to add in the barrier some inward convection, pushing the C ions inwards, but this hinders correct line ratio evaluations. The confinement time ip of the injected elements is always increased when ED is activated. However, this is due to a modification of the edge transport. The core plasma transport (diffusion coefficients, convection velocities, central reduced transport region extension) is not modified when the ED is activated.

3 . INTRODUCTION The ergodic divertor (ED) experiment is a major tool in the physics program of Tore Supra (TS) [], the first Tokamak in which the ED concept has been integrated in the machine design. The ED configuration consists of six, toroidally equally spaced, octopolar coils installed inside the vacuum vessel. Their toroidal and poloidal extensions are 3 and 20, respectively. Their toroidal and poloidal mode numbers are n=6 and m=8 ± 3 (i.e., the system is resonant for a safety factor at the edge qa = 3), thus providing an ergodic layer well located at the plasma edge (with a thickness 0-20 % of the plasma radius), with a rapid spatial decay of the magnetic perturbation towards the plasma core [2]. When activated, the ED strongly modifies the plasma edge [3], resulting in capabilities to screen impurities [4], to stabilise the m=2, n=l tearing mode [5], to reduce the turbulence level [6] and to enhance the radiated power fraction in radiative layer control experiments [7]. Moreover, supplementary heating pulses (both ion cyclotron resonance (ICRH) and lower hybrid (LH)) in the 3-5 MW power range have been coupled to ED plasmas [7,8]. In this paper intensity ratios of H-like and He-like Carbon ion lines are first analysed using an impurity transport code. The three following lines are considered: C VI 33.74A Lyman-aline (Lya), C V 40.27A Is 2 -ls2p lso -!Pl and C V 40.73A Is 2 -ls2p *So - l?3 lines. The two C V lines are called, respectively, resonance R and intercombination I lines and their ratio R I (the G-ratio) is, at Tokamak electron densities, an inverse function of the electron temperature Te [9]. It is concluded, based on the successful simulation of Lya R that the Carbon ion diffusion coefficient increases in the edge ergodic layer (a fact already mentioned in Ref. [4]). However, it is not possible to simulate simultaneously the C V G-ratio without considering charge exchange recombination (CX) of H-like C ions with neutral hydrogen isotopes. With the ED activated, laser blow-off injections of Nickel ions have clearly shown an increase, by a factor of about two, of the central injected

4 impurity confinement time Tp, both for ohmic and ICRH-heated plasmas. Simulations, performed as reported previously for both JET and TS plasmas [0,], require the introduction of a peripheral transport barrier (with reduced diffusion andor increased inward convection) in the same peripheral region where the turbulence level is reduced [6]. A simple local decrease of diffusion is compatible with the C ion line ratios, but is insufficient for successful laser blow-off injection simulations. Inward convection must be increased, but this modifies the LyocR ratio to a value not in agreement with experiments. This paper is organised as follows. Section 2 contains a description of the experimental conditions. The simulation code is shortly reviewed in Section 3. At the same time the procedure followed to describe the sawtooth discontinuities in the quiescent plasma transport is recalled [0,]. Section 4 analyses the two intensity ratios of H-like and He-like Carbon ion lines. In Section 5 a laser blow-off injection into an ED plasma is extensively simulated, with a similar injection into an ohmic plasma treated subsequently for comparison. The consequences of the necessary transport parameters on the line ratios of Section 4 are also discussed. For completeness, before the conclusions of Section 7, two Ni injections into ICRH-heated plasmas with and without ED are rapidly presented in Section 6. 2 DESCRIPTION OF THE EXPERIMENT AND EXPERIMENTAL RESULTS TS is a circular superconducting Tokamak with major radius R = m, limiter radius al = m, toroidal magnetic field Bj = 3-4 T, plasma current Ip = -.5 MA. ICRH and LH supplementary heatings with injected powers in the 3-5 MW range are routinely available. The ED can be activated either during the entire discharge duration or for a few seconds during the Ip plateau. In the former case ED-related effects appear when, with increasing Ip, the qa = 3 resonance condition is attained. Similarly, in the latter case the Ip plateau value is set at the required value to satisfy the resonance condition. Intrinsic impurity line brightnesses in the XUV range (0-300 A) have been recorded using an extreme grazing incidence vacuum spectrometer equipped with two microchannel plate detectors [2]. For all the reported experiments the impurity content is largely dominated by Carbon, with negligible amounts of O, Cl, Fe and Cu.

5 Impurities are injected at a bottom diagnostic port by the laser blowoff technique. The injection system, based on the laser vaporization technique [3], is an improved version of the old TFR Tokamak system [4]. The injected trace atoms reaching the plasma periphery are in the 07 range. Only a fraction reaches the plasma centre, where trace ion densities of the order of -2 0*6 m~3 are acceptable (i.e. non-perturbing) in all plasma regimes except those too close to the critical density. Injected element VUV spectroscopic lines are detected, typically with 2-4 ms time resolution, 60 toroidally away from the injector. A grazing incidence VUV duochromator equipped with two metallic multi-anode microchannel plate detectors is used [5]. 0 anodes spaced by.3 A allow simultaneous detection of several lines such as the strong An=0 transitions of the Li-like, Be-like, Na-like and Mg-like isoelectronic sequences (this is possible because of the higher order sensitivity of the holographic grating). For the reported measurements the line of sight is central, Nickel is the injected element and the Ni XVIII 292A and Ni XXV 8A lines are detected simultaneously in the second and fifth order, respectively. The progression of the injected impurities is also followed with good spatial and temporal resolution by a soft X-ray camera. For the reported measurements only the system with 44 viewing lines in a vertically oriented fan is used. The spacing of the viewing lines in the equatorial plane is about 30 mm, a 00 jim Be filter helping to discriminate the injected impurity emission from background plasma emission. Since only "quiescent" transport is studied, a time resolution in the millisecond range is sufficient. TS having circular plasmas, the horizontal viewing lines are not essential, the radius dependent emissivities Esx being obtained by Abel inversion of the vertical brightnesses only, taking into account the inner plasma eccentricity (Shafranov's shift). Intrinsic carbon and injected nickel are studied in a series of discharges having a 6 s,.4 MA current plateau. The plasma radius is R=2.4 m, the limiter radius al=0.75 m, the toroidal magnetic field BT=3.6 T, the base gas deuterium, but with a non negligible helium content (because of the discharge cleaning procedure). The ED current (led) is switched on at its maximum value (i.e., at maximum magnetic field perturbation), approximately in the middle of the current plateau, after steady state conditions are reached, and then sustained until the end of the current plateau. Nickel is injected during the ED current pulse, as it appears clearly for a typical shot at t=7.5 s in Fig. (shot number #4343), where the currents Ip and IDE, the central electron density n e (0) and temperature 4

6 Te(O) (both from Thomson scattering) are presented, along with a central soft X-ray channel signal BsX- Fig. 2 shows the C line brightnesses detected by the XUV spectrometer, including the C VI Lya line, the two C V R and I lines and the C IV 245A line. All time evolutions are typical for ED plasmas [4], Both Lya and the R line decrease with the ED, but their ratio changes from about.8 to about 0.9. On the other hand, the I line is roughly constant, implying that the G-ratio (ratio of the I line to the R line) increases from about 0.4 to about 0.6. The large increase of the peripheral C IV line with ED is in qualitative agreement with bolometric maesurements [4], showing during ED activation a poloidal asymmetry (located on the plasma low field side, as the ED coils). This suggests the same poloidal asymmetry for the increase of the C IV line (no brightness poloidal profile being available for this line). On the other hand, the radial profiles of the emissivities of Lya and of the R line have the expected shape for poloidally homogeneous emitting shells [4], thus justifying the use of 'central' C VI Lya for the evaluation of the central C content. Proceeding as described in Ref. [6] (i.e., coupling visible bremsstrahlung emission, XUV spectroscopy, and soft X-ray data), the central C density is estimated to be with ED of the order of 07 m~3 ( a reduction of approximately a factor of three from its non ED value). 'Central' lines of O and Cl behave like C VI Lya when the ED is activated (i.e., their brightnesses decrease, implying plasma purification [4]). The same has been observed for Cu lines, occasionally present during LH pulses [8]. It is, however, worthwhile to recall [8] that Cu bursts (probably produced at the LH waveguide mouth) seem 'unscreened' by the ED, similarly to the laser blow-off Ni injections discussed below. Fig. 3 compares the Ni injection during ED of #4343 (solid line) with the one without ED of #4346 (dashed line), showing the Ni XVIII and Ni XXV line brightnesses along with a central soft X-ray signal. It is clear that the decay is much longer when the ED is activated. The decay times of the 'central' brightnesses are identified with the impurity confinement time xp and are often used to compare different discharge regimes. Fig. 3 shows that ip increases when the ED is activated. This is a general finding in TS, also shown in Figs 4 and 5. The former presents two discharges with (#3909) and without (#374) ED, but both in presence of ICRH. The latter shows the tp-values experimentally found on TS as a function of the total input power Ptot, including both ohmic and supplementary (either ICRH or LH) heatings. The circles and the stars are, respectively, for injections with ED on or off. In both cases (with and without ED) confinement degradation appears when increasing Ptot above about 2MW. 5

7 Given the intrinsic impurity 'purification' following ED activation, it is quite surprising that the injected Ni ions reach the central plasma, apparently unaffected by the presence of the peripheral ergodic layer. Waiting for more detailed studies, at the present time the only possible explanation for this unexpected behaviour is a local destruction of the peripheral shielding connected with the large local perturbation caused by the injected puff []. 3 IMPURITY MODELLING The impurity transport simulation code describes, for a given atomic species, in cylindrical geometry, ionisation,recombination and radial transport of the ions of charge z and density nz [7]. The procedure followed in laser blow-off experiments to compare simulations with the relevant diagnostic data has been described in detail in Ref.[0]. In the Appendix of Ref.[ll] the atomic data needed to describe ionisation, recombination and both line and continuum emission are reviewed. Charge exchange (CX) recombination with neutral hydrogen isotopes is also included. The impurity flux density Tz is expressed as the sum of both diffusive and convective terms r z (r) = - D(r) Vn z (r) - V(r) n z (r) () where D(r) and V(r) are radius dependent diffusion coefficients and inward convection velocities, both taken independent of the charge z of the ions. Taking V(r)=(ra)VA, where a is the last mesh radius, a few centimetres outside the last closed flux surface (LCFS) (for exemple, for #4343, a=0.78m, whereas al=0.75 m), and assuming both D and VA to be radially constant, yields at steady state a Gaussian shape profile for the total impurity density nt(r) = nt(o) exp(-s r2a2),with a peaking parameter S= ava 2D. An impurity confinement time xs for particle transport along the field lines can also be added in the peripheral region. Electron density ne(r) and temperature Te(r) profiles are needed to evaluate the required atomic physics rates. The former are obtained by adjusting the spatially resolved vertical Thomson scattering data to the five vertical IR interferometer channels. The average T e (r) profile is obtained by Thomson scattering, the effect of sawtooth modulations being given by the

8 ECE emission. The sawtooth period is divided discontinuously into three parts, each with different Te(r) profiles. Since from the point of view of transport the plasma core can be divided into two distinct regions [0,], the starting point of the simulation takes D(r) constant in each region with a third transition region in-between. D(r) is subsequently adjusted, generally smoothing the transition and extending or reducing the width of the central region, to follow the absolute values of the soft X-ray emissivity profiles Esx(r). V(r), on the other hand, is initially obtained by supposing a peaking factor S=l everywhere, i.e. we take V(r)=(ra) VA(r) = 2r D(r)a2. This implies a depression towards the centre of the straight line characteristic of the constant D-VA approximation. Physically, this is necessary to avoid the too large simulated values of the central emissivities Esx(O) obtained with large central peaking factors S. During the inflow phase diffusion dominates and the simulation is not much sensitive to the exact shape of V(r) at the transition region and in the inner half of the large diffusivity region. This is somewhat different from the gradually diffusing JET H-mode injections reported in Ref.[8], where the slowly increasing Esx(r) peak amplitudes could be controlled by the V(r) function at and just outside the wide transition region. In the absence of sawteeth, the injected impurities reach progressively the centre, Esx(O) increases regularly, until Esx(r) is not anymore hollow. The central V(r) depression is quite sensitive to the details of the evolution of the Esx(r) profiles during this phase. Subsequently, when Esx(O) starts to decrease, the depletion phase begins. During this phase the decay of the Esx(r) profiles is sensitive to the ratio V(r)D(r), and consequently V(r) can be better determined. This is, however, possible only for sufficiently long quiescent phases between sawteeth, which is not the case for the four simulations presented in the following. The central V(r) depression is general, except in the pellet enhanced performance (PEP) regime, where it reverses (VA(r=>0) is larger than VA(a), whereas D(r) has the usual two transport region shape) [9]. Simulation of the peripheral line brightnesses is a well-known difficulty in laser blow-off experiments [0,20,2]. It must be remembered that spectroscopic lines are generally detected toroidally away from the laser blow-off injection port. To the best of our knowledge, only Kocsis et al. [22] have observed the injected Al particles also in the injection port, showing that the interaction of the blow-off puff with the peripheral plasma is rather complex. In the simulations, the unknown neutral impurity influx O(t) at the plasma edge is generally tailored to follow the relative line brightnesses of 7

9 the most peripheral ion monitored (e.g., Ni XVIII 292A in the present experiments). The brightnesses of these low ionisation potential ions show a rapid burn-through ionisation peak, followed by a tail becoming more important when the emitting shells are located more deeply. It is also wellknown that their position depends on the peripheral Te profile. In the first TFR measurements the tail slopes have been interpreted as the peripheral impurity confinement time [23], but it was not possible to simulate them in the frame of the constant D-VA approximation [20]. Indeed it was necessary to use "ad-hoc" <I>(t) functions, having a triangular shape followed by a low exponential tail a few tens of milliseconds long. This procedure was adopted subsequently for the JET and TS simulations [0,] and was considered to reproduce accurately the time evolution of the total influx of impurities just within the LCFS. It was, however, difficult to justify the temporal shape of O(t) as a product of laser ablation at the target and it was thought that unknown phenomena during the first ionisations and during the rapid toroidal dispersion of the locally injected puff determine its width. Moreover, recent JET results, modelling the line brightnesses of several non-central Fe ions, have evidenced the existence of an edge impurity transport barrier [24]. In view of all these considerations, in the present simulations O(t) is taken triangular with a width of only -2 ms, determined by the experimental evolution of the Ni XVIII brightness during the first milliseconds. The experimental Ni XVIII tail is, therefore, determined by 'capture' or confinement phenomena in the peripheral plasma. To show qualitatively the latter phenomena, the simulations of the brightnesses of Ni XI and Ni XVII (Ar-like and Mg-like ion lines) will be presented along with that of the brightness of Ni XVIII (Na-like ion line), available experimentally. The simulations presented in Section 5, even if they fail to simulate satisfactorily the Ni XVIII line brightness, confirm nevertheless that peripheral confinement determines the low ionisation potential ion tails, confirming consequently the old interpretation of Ref.[23]. The soft X-ray emissivity profiles exhibit discontinuities at sawtooth crashes that in most cases cannot be simulated by considering the drop of Te alone [0]. At crash, the impurity transport must be greatly increased over the central region (having a radius roughly 40% wider than the Esx(r) inversion radius rinv), leading to a major redistribution of the impurity density. Following the ideas developed in Ref.[25], in Ref.[0] the modifications of the ion density radial disribution at crash were described heuristically by increasing, with respect to the quiscent phases, both D(r) and V(r) by orders of magnitude during a crash period lasting us. This 8

10 procedure is followed for the present TS simulations. A unified description is not possible, since ad-hoc adjustement of the transport perturbation is necessary for each experimental situation. Anyhow, the movement of impurities during a crash is in most cases more complicated than a simple flattening of the impurity distribution [26]. If the sawtooth period is shorter than approximately 50 ms, the soft X-ray central brightnesses Bsx(t) show that the first sawtooth after injection is always inverted (Bsx(t) increases at crash), i.e., impurities are pushed towards the center discontinously at the crash [0]. Inverted sawteeth can be simulated either by increasing D(r) alone or by increasing both D(r) and V(r) (the latter only around and outside rinv) [0]. During the decay phase the central Bsx(t) sawtooth modulations are normal (Bsx(t) decreases at crash). For the simulations reported here, it is sufficient to increase only D(r), even if expulsion of the injected ions from the central region cannot be excluded, for example in the presence of ICRH. This expulsion can be simulated by a central enhanced (by orders of magnitude) outward (i.e., negative) V(r) function [7]. Note that in all cases the transport perturbation extends up to a radius about.4 times greater than the sawtooth inversion radius rinv. It is worthwhile also to stress that the adopted ad-hoc procedure is based on the simulation of both kinds of sawteeth through a one-dimensional ion movement. However, in one simulation reported in Ref.[ll], this hypothesis has been questioned on the basis of a discrepancy at sawteeth between experiment and simulation for the Ni XXV line with a hollow emissivity profile peaking around 4 NUMERICAL SIMULATIONS OF CARBON LINE INTENSITY RATIOS In Table the results of different simulations of #4343 with the ED on (n e (0) m" 3, T e (0) = 2.4 kev) are presented. The n e (r) and Te(r) profiles have been obtained as previously described. However, the peripheral extrapolation for ral > 0.8 is important for the C line ratios. The n e (r) extrapolation is in agreement with microwave reflectometry in the equatorial plane. Concerning the peripheral Te(r), Thomson scattering values are available at ral * 0.8 and at ral =0.9, whereas reciprocating Langmuir probes give Te-values around and beyond al. The chosen Te(r) profile is smooth at the periphery, without the discontinuities suggested by the theory for ral-values between 0.8 and about 0.9 [6]. Carbon line intensities have been calculated as described in Refs [ and 27]. It is worthwhile to recall

11 that the CX contribution to the I line is evaluated by supposing that three quarters of the recombining collisions of C5+ ions populate excited triplet states and that deexcitation to the ground state is only possible through the I line. The simulations are for steady state conditions. Therefore the carbon ion distribution depends basically on the peaking factor S and is approximately independent of the presence of a reduced transport inner region, as long as S is roughly radially constant. In the first seven simulations the diffusion coefficient and the peaking factor are taken constant and equal, respectively, to Dout=l-5 m^s and to S=l. D e rg is the diffusion coefficient (larger than Dout) between 0.65 m and a=0.78 m. In the xs column 'no' means xs equal to infinity (i.e., no parallel flow), 'yes' means TS=.5 ms between 0.72 m and 0.78 m and is=3 ms between 0.65 m and 0.72 m. In the nn column 'yes' a nd 'no' specify the inclusion of a neutral density profile nn(r). For its choice we are guided by the interpretation of the JET C Lyman series spectra [27]. For r=a, nn is taken of the order of 0*7 m~3, decreasing to about 06 m-3 at r=0.68 m, i.e., at the radius for which in the reference simulation the H-like and He-like C ion densities are approximately equal. It has been verified that simulation (not very sensitive to the ne(r) and T e (r) profiles) is quite satisfactory for the pre-ed LyaR and G ratios, shown in Fig.2 to be equal to about.8 and 0.4, respectively. On the other hand, to obtain LyaR = during ED, it is necessary (simulations 2-4) to increase either perpendicular or parallel (or both) peripheral transports in the ergodic edge region. Increased transport has already been considered to simulate the increased peripheral C IV and C III lines when the ED is activated [4]. The width of the enhanced transport edge region (from r=0.65 m to r=a=0.78 m) is the minimum width required to reduce the LyaR ratio from almost two down to one. However, the perpendicular transport enhancement (simulations 2-3) displaces inwards the C V ion emitting shells (i.e., toward larger Te-values). The G-ratio, an inverse function of T e [9], thus decreases from 0.36 to Increasing only parallel transport (simulation 4) does not affect the G-ratio, which is nevertheless much smaller than the experimental value. To obtain G «0.6 it is necessary (simulation 5) to enhance the I line through CX processes by including in the edge region large neutral densities. Simulations 6 and 7 show that both increased transport and large nn-values are necessary to obtain satisfactory simulated line ratios. 0

12 Before describing the Ni injection simulations, it must be said that, if one takes a T e (r) profile with a large gradient between al (as discussed in Ref.[6]), the agreement of the two C line ratios with the experiment is not as good as when one takes a smooth T e (r) profile. The simulated variations at the ED activation are roughly halved with a nonsmooth Te(r) profile. TABLE SIMULATION OF THE CARBON LINE INTENSITY RATIOS LyaR AND G (when the ED is on, their experimental values are =0.9 and ~ 0.6, respectively) Simulation LyaR G DergDout see Fig.9 see Fig. 0 ts no no yes yes no yes no no no n n no no no no yes yes yes yes yes 5 SIMULATION OF TWO OHMIC NI INJECTIONS (WITH AND WITHOUT ED) Shot #4343 (already presented in Figs. -3 and also considered for the line intensity ratio simulations of Section 4) is first considered, starting with the simulation shown in Fig.6, performed with the assumption of two regions having different transport parameters [0]. This first simulation is not satisfactory, as far as the inflow phase is concerned, and other simulations are necessary. A good one, as far as the core plasma is concerned, is presented in Fig.7. In Fig.8 Ni XVIII is optimised, but the peripheral transport is in disagreement with the conclusion of Section 4. The two successive simulations (Figs 9 and 0) take into account the conclusions of Garbet et al. [6] on peripheral transport.

13 Going back to Fig.6, the three signals of Fig.3 are presented in the upper part (from top to bottom), i.e., the Ni XVIII and Ni XXV lines and a central soft X-ray brightness. Solid lines show the simulated signals, to be compared with the dashed lines showing the experimental ones. For Ni XVIII the brightnesses Bnor are normalized at maximum, whereas for Ni XXV and the soft X-ray signals normalization is on the decay phase. Along with Ni XVIII, the simulated Ni XI 48A and Ni XVII 249 A brightnesses are shown (dotted and dot-dashed lines, respectively, normalized for clearness at 0.5) to give a visual idea of the peripheral 'capture' or confinement phenomena. The lower part of Fig.6 shows the assumed D(r) and V(r) profiles (respectively, solid line in m^s and dashed line in ms with a constant multiplicative factor), implying the assumption of two regions with distinct transport properties with a third transition region in-between. As discussed in Section 3 and similarly to Refs [0,], their shapes have been obtained by following the inverted Esx(r) profiles. Fig.6 shows that it is possible to simulate the Ni XVIII line, but only the decays of the Ni XXV line and of the central soft X-ray brightness. The peripheral 'capture' is correctly simulated, but the inflow phase towards the centre is not at all satisfactory. It is therefore necessary to modify the D(r) and V(r) curves in such a way as to allow a correct fast inflow, while preserving the peripheral 'capture'. Several simulations are presented in the following for #4343 and for the reference injection without ED (#4346). All the figures are presented in the same way as Fig.6. Fig.7 shows a good simulation obtained by increasing Dout from about 0.5 m 2 s up to about 2 m 2 s in the outer region. The inflow is faster and the required decay is recovered with a peripheral V-barrier (V of the order of 00 ms around r = m). However, the peripheral 'capture' is only partially simulated. Given the few experimental data available for the peripheral region, a detailed study of the sensitivity of both width and position of the peripheral V-barrier is not possible. For variations not too large compared with the case shown in Fig.6 it is always possible to recover a satisfactory simulation by varying slightly Dout (up to 0% at maximum). Moreover, it is not possible to follow the very fast signal rises of the soft X- ray brightnesses, but this is general for all simulations reported in this paper. It appears therefore that equation () does not describe satisfactorily, at least in TS, the first 5-20 milliseconds after the injection []. With the purpose of optimising Ni XVIII, the best result is shown in Fig.8. With respect to the case of Fig.7 the peripheral transport barrier is now a superposition of both decreased convection and decreased diffusion 2

14 (both by a factor of two). It was necessary to reduce D from the interior limit of the V-barrier up to the last mesh radius, but this is in strong disagreement with the conclusions of Section 4 that Derg is larger than D O ut- C02 laser scattering and microwave reflectometry maesurements have shown the existence of a low transport parameter region around al [6], in the intermediate region between the plasma bulk and the ergodic zone, where the radial electric field changes its sign when the ED is on. The measured density fluctuations are usually connected with diffusivity. Moreover, power balance analysis of the same discharges has evidenced in the same region a reduced thermal diffusivity xe- Guided by these findings, the simulation presented in Fig.9 is obtained by reducing D (by a factor of about ten) in the intermediate region and increasing it in the exterior ergodic region. In this way it is possible to satisfy the C line intensity ratio constraints, as shown by simulation 8 of Table (obtained with the same transport parameters as for Fig.9). However, Fig.9 shows that the simulation of the blow-off decay is too rapid. To lengthen it, it is necessary to increase V (always in the al region) and this is shown in Fig.0. This simulation is quite similar to the one presented in Fig.7 but simulation 9 of Table (performed with the same transport parameters) does not agree with the experiment as far as the LycxR ratio is concerned. This ratio goes up to 2.3, since the increased convection acts outwards of the H-like ion emitting shells and pushes the C ions inwards. Our conclusion is that for shot #4343 (with ED on) it not possible to simulate simultaneously both C line intensity ratios and Ni laser blow-off injection in the frame of the simple flux expression (). However, a different transport behaviour of C and Ni ions (which have not the same origin) cannot be excluded. To conclude our discussion on #4343, it is worthwhile to notice that, as far as Ni XVIII is concerned, the simulation of Fig.7 is better than that of Fig.0, whereas the simulations of the other two brightnesses are practically identical. Fig. shows a simulation of shot #4346 ( n e (0) = 3.2 0*9 m-3, T e (0) = 2.9 kev) without ED activation. Again, two transport parameter regions are assumed. Only the soft X-ray central brightness is correctly simulated, whereas the Ni XXV simulation is not as good as for shot #4343. It is not possible, within the boundary O(t) condition assumed (i.e., triangular shape -2 ms wide without tail) to simulate the Ni XVIII line, which, incidentally, as evident in Fig.3, has a time evolution not much different from that of the same line for shot #4343. Adding a peripheral V barrier in 3

15 the ms range (i.e., of the same order of magnitude as that needed in Refs[ll,24]), i.e., is about a factor of ten smaller than for shot #4343), modifies only slightly both Ni XVIII and Ni XXV time evolutions. To prove that the assumption of a O(t) tail (put forward in Ref.[20] and used, between others references, in [0,]) is not a good one to explain complicated peripheral 'capture' phenomena by acting artificially on the recycling flux, another simulation of the same shot #4346 is presented in Fig. 2 without changing the transport parameters. The addition of a weak exponential tail to the flux O(t) (only 5% of the maximum with a time constant of 30 ms) is sufficient to simulate correctly both Ni XVIII and Ni XXV, but, at least for this simulation, the simulated Bsx(t) reaches its maximum too late. This is caused by the unphysical 'supplementary' flux reaching the centre. 6 SIMULATIONS OF LASER BLOW-OFF INJECTIONS WITH ICRH HEATING Figs 3 and 4 present the simulations of two injections performed during ICRH heating pulses. The first one (Fig 3) is for shot #3909 (n e (0) = 5,8 0!9 m-3, T e (0) = 2.65 kev, PiCRH = 2 MW) with the ED on, whereas the second one (Fig.4) is for shot #374 ( n e (0) = 3.8 0*9 m-3, T e (0) = 4.3 kev, PiCRH = 2.5 MW) without ED. Both plasmas are in a giant sawtooth regime ( AT e (0) =900 ev instead of AT e (0) =250 ev for ohmic plasmas). Physically the same hypothesis, as discussed in the previous Section for the ohmic plasmas, are necessary. However, both the transition region between the two transport regions and the inner diffusion coefficient Din can be inferred better than in the previous ohmic plasma cases due to the longer sawtooth period. On the other hand, the outflow is more uncertain, as a consequence of the large perturbations associated to the irreproducible giant sawteeth and the related uncertainties on the soft X-ray background subtraction. When the ED is on (Fig. 3), a peripheral V barrier ( always close to 00 ms) is required, as in Section 5. Also for this injection it is not possible to distiguish (as between Figs 7 and 0) the cases of increased V only and of simultaneous reduction of D and increase of V. For the case without ED (Fig.4), contrary to the corresponding ohmic shot of Fig.ll, the inclusion of a V barrier of about 20 ms improves mildly the Ni XVIII and Ni XXV simulations. Therefore it has been preferred to the case without V barrier. 4

16 7 CONCLUSION The main conclusion of this paper is the explanation of the increased injected Ni confinement time xp during ED activation as the effect of a peripheral transport barrier in the al region, in agreement with independent measurements. Garbet et al. [6], from the analysis of CO2 laser scattering and microwave reflectometry data, have inferred the existence of a low transport parameter region in the intermediate region between the plasma bulk and the ergodic zone. These experimental data have been analysed in terms of ne fluctuations, i.e., in terms of diffusivity. Moreover, power balance analysis of the same discharges has evidenced in the same region a reduced thermal diffusivity xe- It is not possible to simulate the experimental xp-values only by reducing the diffusion coefficient in the considered radial region. It is necessary at the same time to increase the inward convection velocity. However, the case of simple increased convection cannot be distiguished from the case of simultaneous decreased diffusion and increased convection. In the case of a pure convective barrier the required convection velocities (about 00 ms) are of the same order of magnitude as those used in the JET H-mode injection simulations [8]. However, the core Dout-values for JET H-modes are smaller than the values reported both for the injections in JET L-mode plasmas and for the injections in TS reported in this paper. In both cases (with and without ED) confinement degradation is observed, when increasing the total input power Ptot- From the point of view of the transport barrier, the ED plus ICRH Ni injections can be simulated in the same way as the ohmic ones. As far as the core plasma transport parameters and the dimensions of the central reduced transport region are concerned, there is not much difference between plasmas with and without ED. This conclusion is in agreement with that of Harris et al. [28], who found that the mean thermal diffusivity %e shows the same n e dependence as that exhibited by similar discharges without ED. As far as injections without ED are concerned, globally the simulations confirm previous JET and TS ones [0,,8,29]. In some cases the introduction of a peripheral convection barrier (with V-values one order of magnitude lower than in the ED case) [,24] improves at least partially the Ni XVIII and Ni XXV line simulations. Unfortunately an extended study by 5

17 varying the plasma parameters in a wide range, as already done on JET [29], has not yet been possible on TS. One has to remark, nevertheless, that the transport transition region includes the sawtooth inversion radius rinv, which in TS is typically around 0.30 m.. It has been proved for a particular Ni injection into an non-ed ohmic plasma that the assumption of a O(t) tail in the boundary incoming Ni flux is not a good one to explain complicated peripheral 'capture' phenomena. A weak exponential tail of the flux O(t) (only 5% of the maximum with a time constant of 30 ms) is sufficient to simulate correctly both Ni XVIII and Ni XXV, but, at least for this simulation, the simulated central soft X-ray brightness reaches its maximum too late. This is caused by the 'supplementary' flux O(t) reaching the centre, which is consequently unphysical. At the present time, from the TS simulations presented here, from the corresponding ones from JET [0,8,29] and TFTR [2], from the measurements in two different ports (including the injection port) [22], it can be stated that the interaction of the injected puff with the peripheral plasma is still a not well understood phenomenon. This is quite surprising if one remembers that the injection system has not changed at all since the first one introduced about twenty years ago [3]. In TS plasmas with the ED activated it is not possible to simulate at the same time both Ni injections and C line Lyoc R and G intensity ratios. To obtain LyocR equal to about one, increased peripheral transport (either parallel or perpendicular, or both simultaneously) from r» 0.65 m up to the last mesh radius at r=0.78 m is required. To recover a G-ratio of about 0.6 it is necessary to introduce CX recombination processes by assuming a neutral density profile decreasing from a value of the order of 07 m~3 at the last mesh radius to about 06 nr 3 at r=0.68 m, where in the absence of CX processes, for the reference case without ED, the H-like and He-like C ion densities are approximately equal. By further adding in the simulations a purely diffusive barrier, in the intermediate region between between the plasma bulk and the ergodic zone, it is possible to keep satisfactory C line ratios, but the simulated confinement time xp is too small. It has therefore been necessary to supplement the diffusive barrier with a convective one. In this case the simulated xp value is satisfactory, but the LyaR ratio becomes too large. 6

18 ACKNOWLEDGEMENTS The authors are grateful to P. Cristofani, Th. Dudok de Wit, J. Lasalle, Y. Margerit, A.L. Pecquet and B. Saoutic for valuable help. They express also their appeciation to the TS operating Team. 7

19 REFERENCES [] GROSMAN A., EVANS T.E., GHENDRIH Ph., et al. J. Nucl. Mater. 76&77 (990)49. [2] GHENDRIH Ph., CAPES H., DEMICHELIS C, et al. Plasma Phys. Controll. Fusion 4 (992) [3] GROSMAN A., GHENDRIH Ph., DEMICHELIS C, et al. J. Nucl. Mater. 97&98 (992) 59. [4] BRETON C, DEMICHELIS C, MATTIOLI M., et al. Nucl. Fusion 3(99) 774. [5] VALLET J.C., POUTCHY L., MOHAMED-BENKADDA M. S., et al. Phys. Rev. Lett. 6(99) [6] GARBET X., et al. to be presented at the 994 IAEA Conference on Controlled Fusion and Plasma Physics. [7] GROSMAN A., MONIER-GARBET P., VALLET J.C., et al. J. Nucl. Mater. (994) to be published (presented at the 994 PSI Conference) [8] GONICHE M., GROSMAN A., GUILHEM D., et al. 20th EPS Conference on Controlled Fusion and Plasma Physics (Lisboa,993) Vol.l7C, Part II, European Physical Society (993) 65. [9] TFR GROUP, DOYLE I.G., and SCHWOB J.L. J. Phys. B : At. Mol. Phys. 5.(982) 83. [0] PASINI D., MATTIOLI M., EDWARDS A.W., et al. Nucl. Fusion 0(990)2049. [] MATTIOLI M., DEMICHELIS C, MONIER-GARBET P., et al."laser Blowoff Injection of Metal Impurities in Tore Supra", (993) Rep. EUR-CEA-FC 49 CEN Cadarache (unpublished). [2] SCHWOB J.L., WOUTERS A.W., SUCKEVER S., and FINKENTHAL M. Rev. Sci. Insrum. 5 (987) 60. [3] MARMAR E.S., CECCHI J.I., and COHEN S. Rev. Sci. Instrum. 5 (975)49. [4] BRETON C, DEMICHELIS C, HECQ.W. and MATTIOLI M. Rev. Phys. Appl. 5_ (980) 93. [5] DEMICHELIS C, MATTIOLI M., MONIER-GARBET P., et al. Meas. Sci. Technol. 4 (993) 09. [6] GUIRLET R., MATTIOLI M., DEMICHELIS C, et al. 2th EPS Conference on Controlled Fusion and Plasma Physics (Montpellier,994) to be published. [7] TFR GROUP, Nucl. Fusion 25. (985) 98. [8] PASINI D., GIANNELLA R., LAURO TARONI L, et al. Plasma Phys. Controll. Fusion 4(992) 677. [9] GIANNELLA R., HAWKES N., LAURO TARONI L, et al. Plasma Phys. Controll. Fusion 3JL( 992) 687. [20] TFR GROUP, Nucl. Fusion 2 (983)

20 [2] STRATTON B. C, FONCK R. J., HULSE R., et al. Nucl. Fusion 23 (989) 437. [22] KOCSIS G., BURGER G., IGNACZ P.N., et al. Plasma Phys. Controll. Fusion 24( 992) 423. [23] TFR GROUP, Phys. Letts. 87A (982) 69. [24] DENNE-HINNOV B., GIANNELLA R., LAURO-TARONI L, et al. 2th EPS Conference on Controlled Fusion and Plasma Physics (Lisboa,993) Vol 7C, Part I, European Physical Society (993) 55. [25] COMPANT LA FONTAINE A., DUBOIS M., PECQUET A.L., et al. Plasma Phys. Controll. Fusion 27 (985) 229 [26] PETRASSO R., SEGUIN F., LOTER N., et al. Phys. Rev. Lett. 49.(982)826 [27] MATTIOLI M., PEACOCK N.J., SUMMERS H.P., et al. Phys Rev. A 4Q (989)3886. [27] TFR GROUP, DOYLE I.G., and SCHWOB J.L. J. Phys. B : At. Mol. Phys. 5.(982) 83. [28] HARRIS G.R., CAPES H., and GARBET X. Nucl. Fusion 22.(992) 967. [29] GIANNELLA R., LAURO-TARONI L, MATTIOLI M., et al. Nucl. Fusion (994) in press, (also Rep. JET-P(93)8 ) 9

21 FIGURE CAPTIONS Figure : From top to bottom: time evolutions of the plasma current Ip, the ED current IED, the central electron density ne(0), the central electron temperature Te(0) (both from Thomson scattering) and of a central soft X-ray signal BSX- TS shot #4343. Figure 2 : From left to right and from top to bottom: time evolutions of a few Carbon line brightnesses (count number cts), i.e., C VI Lya, C V R+I, C IV 245A line. Also shown (bottom) the time evolutions of the line ratios r of LyaR and of IR (the G-ratio). TS shot # Figure 3 : From top to bottom: time evolutions of the normalized brightnesses B n or of the Ni XVIII 292A line, of the Ni XXV 8A line, and of a soft X-ray central signal for #4343 (solid line: injection with ED on) and for #4346 (dashed line: injection into a plasma without ED). The laser firing time is taken as time origin. Figure 4 : Same as Fig.3 for two Ni injections into plasmas heated by a ICHR pulse (solid line: # 3909 with ED and PlCRH = 2 MW, dashed line: #374 without ED and with PlCRH = 2.5 MW). Figure 5 : Impurity confinement time xp ( taup in ms) for laser blow-off injections into TS as function of the total power Ptot (ohmic plus supplementary ICRH or LH heating powers in MW). Circles and stars are, respectively, for injections with the ED on or off. Figure 6 : In the upper part time evolutions of the normalized brightnesses Bnor of the simulated signals (solid lines) and of the experimental signals (dashed lines) for #4343 with ED on. From top to bottom the Ni XVIII and Ni XXV lines and a central soft X-ray brightness are shown. Along with Ni XVIII the simulated brightnesses of Ni XI and Ni XVII (dotted and dot-dashed lines,respectively) are shown to have a qualitative indication of the peripheral 'capture' phenomena. The laser firing time is taken as time origin. In the lower part the profiles of the transport parameters D (in m^s) and V (in ms) are given, the latter with a multiplicative factor. Simulation without a peripheral transport barrier. Figure 7 : Same as Fig.6, but with a peripheral transport barrier (see the D(r) and V(r) curves at the bottom). Only the convection velocity is increased with respect to Fig.6. Figure 8 : Same as Fig.6. With respect to Fig.7, D(r) is decreased at the periphery and V(r) is adjusted to have a good xp-value. This allows correct simulation of the Ni XVIII line, but then the simulated C line ratios are not correct. Figure 9 : Same as Fig.6. With respect to Fig.7 the peripheral barrier is obtained reducing D(r) around al- To simulate the C line ratios, D(r) must be larger than Dout at the exterior ergodic edge. 20

22 Figure 0 : Same as Fig.6. With respect to Fig.9 in the peripheral barrier V(r) is increased to recover a correctly simulated xp-value. Figure : Same as Fig.6 for injection into shot #4346 without ED. Figure 2 : Same as Fig., but an exponential tail is added to the incoming flux O(t) to simulate correctly the Ni XVIII and Ni XXV line brightnesses. The inflow phase of the soft X-ray signal is not satisfactorily simulated. The D(r) and V(r) curves are the same as forfig.ll. Figure 3 : Same as Fig.6 for #3909 with ED and with a 2MW ICRH pulse. Figure 4 : Same as Fig.6 for #374 without ED and with a 2.5 MW ICRH pulse. 2

23 .5 < S '400 X m t(s) Figure

24 Figure 2

25 0.8 Ni XVIII ifti( 0.8 g0.6 NiXXV iuu Jl «M- <. ^, K I soft X-rays LL4^ 0.5 Figure 3

26 0.8 Ni XVIII g0.6 NiXXV soft X-rays t(s) Figure 4

27 E * * *. o o ooo QOO -Sgt Jt * t* * * * * 4 o o m (SUl)

28 :nnr Ni XVIII NiXI Ni XVII go.6 CD n \ \ ' r ; ' J i i i i ' ^J J J '»> i <j i t f I! I I T NiXXV i i i I i i soft X-rays i i t(s) \ V" - _ - - ifi s Q r r(m) if Figure 6

29 NiXXV soft X-rays CM o 6.5 * > k Q 0.5h r(m) f ^ t i... \ \ \ ' \ : \ \.. \ _ l I '. * - Figure

30 i i r i r ^rzrz Ni XVIII NiXI NiXXV soft X-rays I " I i i r 2-3 O O.5 * > Q \ r(m) f' -<* ".V... I....,. f ' N ; \ ; i : \ ; I Figure 8

31 i r =-T3 Ni XVIII NiXI soft X-rays ' "~ ^ LO 5 d 4- * >3 Q 2 I L. " " =~J- *tr. -~~~~~ r(m) Figure 9

32 J \ - < \\ \ \ i ^ i - i i _ v y j h Ni XVIII NiXI Ni XVII N ^ ^ A \ ' - -, ^ y -j \ -j v i - - T p NiXXV soft X-rays ! C<j5- : " : II Q 2 ' M... i... u... I t r(m) Figure 0

33 r-r-r Ni XVIII NiXXV if) O * > o '. - ' y ' ' " ' _' _, -^- _ r(m) Figure

34 =~r-r Ni XVIII NiXI NiXVII NiXXV 0.25 if) O >.5-. jr. Q 0.5h : ^ ^ ^ ^. i r r(m) Figure 2

35 NiXXV C\J q 2 o *.5 Q : r(m) U H t...:.] r - i I I t " \ i Figure 3

36 Ni XVIII NiXI NiXVII One C 0.6 CD NiXXV soft X-rays ^N > ^ *3- o * >2 Q - _ t(s) -^" fir JU r* """ "" " r(m) r I : I ; ; : ; : : \ J i I I i \ _ Figure 4