Pedestal and Er profile evolution during an edge localized mode cycle at ASDEX Upgrade

Pedestal and Erprofile evolution during an edge localized mode cycle at ASDEX Upgrade M. Cavedon1, T. P¨utterich1, E. Viezzer2, F. M. Laggner3, A. Burckhart1, M. Dunne1, R. Fischer1, A. Lebschy1,4, F. Mink1,4, U. Stroth1,4, M. Willensdorfer1, E. Wolfrum1and the ASDEX Upgrade Team 1Max-Plank-Institut f¨ur Plasmaphysik, D-85748, Garching, Germany 2Dept. of Atomic, Molecular and Nuclear Physics, University of Seville, Avda. Reina Mercedes, 41012 Seville, Spain 3Institute of Applied Physics, TU Wien, Fusion@¨ OAW, 1040 Vienna, Austria 4Physik-Department E28, Technische Universit¨at M¨unchen, 85748 Garching, Germany 18 August 2017 Abstract. The upgrade of the edge charge exchange recombination spectroscopy diagnostic at ASDEX Upgrade (AUG) has enabled highly spatially resolved measurements of the impurity ion dynamics during an edge-localized mode cycle (ELM) with unprecedented temporal resolution, i.e. 65 µs. The increase of transport during an ELM induces a relaxation of the ion, electron edge gradients in impurity density and flows. Detailed characterization of the recovery of the edge temperature gradients reveals a difference in the ion and electron channel: the maximum ion temperature gradient ∇Tiis reestablished on similar timescales as ∇ne, which is faster than the recovery of ∇Te. After the clamping of the maximum gradient, Tiand Teat the pedestal top continue to rise up to the next ELM while nestays constant which means that the temperature pedestal and the resulting pedestal pressure widen until the next ELM. The edge radial electric field Erat the ELM crash is found to reduce to typical L-mode values and its maximum recovers to its pre-ELM conditions on a similar time scale as for neand Ti. Within the uncertainties, the measurements of Eralign with their neoclassical predictions Er,neo for most of the ELM cycle, thus indicating that Eris dominated by collisional processes. However, between 2 and 4 ms after the ELM crash, other contributions to E×Bflow , e.g. zonal flows or ion orbit effects, could not be excluded within the uncertainties. 1. Introduction Since the discovery of the high confinement mode, an accompanying edge instability, called edge-localized mode (ELM), has been observed. ELMs pose one of the most serious technical challenges for future devices due to the expulsion of energy and particles onto plasma facing components within about 1 ms. Moreover, the fusion performance Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 2 projections for ITER are predicted to be bound to the pedestal characteristics [1, 2]. It is, therefore, crucial to study the evolution and recovery of the edge profiles throughout the entire ELM cycle. In particular, the dynamics of the E×Bvelocity in the edge region is of fundamental interest since its shear is believed to play an important role in the reduction of the edge turbulent transport [3]. While the recovery of the electron density (ne) profile and the electron temperature (Te) profile after an ELM-crash have been investigated in detail [4, 5, 6, 7], the evolution of the impurity flows and the ion temperature at the plasma edge are only poorly documented so far due to limited measurements capability. Previous studies indicate that the response to the ELM onset is a rapid increase in the temperature and E×B velocity in the steep gradient region followed by a drop with a longer timescale [8]. However, a comparison between the recovery of the electron and ion profiles was not performed due to the limited time resolution (550 µs) in [8]. A faster recovery of the ion temperature compared to the electron temperature was suggested in [9] but, so far, not experimentally verified. Similar studies were also performed at ASDEX Upgrade (AUG) where a collapse of the edge Erwas observed at the ELM crash [10, 11]. Here, however, the time resolution was limited to 2.3 ms. At the MAST tokamak, the dynamic evolution of the edge Erprofile during the ELM cycle was studied with a time resolution of 200 µs [12]. The edge radial electric field is observed to recover within 2 ms after the ELM crash while the electron and ion profiles could not be compared. In this work, using the new set of edge diagnostics installed at ASDEX Upgrade (see section 2), a full characterization of the profile evolution of the edge density, temperatures, poloidal and toroidal flows with a time resolution of 65 µs is presented. The dynamics of the ions and electrons are compared and the evolution of helium as a trace impurity is investigated. Moreover, a comparison between the measured radial electric field and its neoclassical prediction during the ELM cycle is presented. 2. Experimental setup and methodology In the last years, AUG has been equipped with several diagnostics that greatly improve the temporal and the radial resolution of measurements covering the plasma pedestal. These diagnostics as well as new analysis methods are presented following section. 2.1. Electron measurements In the present work, the edge profiles of electron density neand temperature Tehave been obtained from Lithium Beam Emission Spectroscopy (Li-BES) [13, 14, 15] and the Electron Cyclotron Emission (ECE) radiometers [16], respectively. The diagnostic measurements have been combined via the integrated data analysis algorithm (IDA) [17] which, in the framework of the Bayesian theory, reconstructs the neand Teprofiles and their gradients with a time resolution of about 100 µs. Moreover, the ECE forward modelling developed in [18] has been employed to model the ECE emission and hence to Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 3 Figure 1: New method for the evaluation of intra-ELM neprofiles based on Li-BES diagnostic: (a) evolution of the emission of one line of sight (LOS) at ρ= 0.993, (b) divertor shunt current Idiv, (c) synchronized time traces of the measured signal (red) and passive (blue) and active (green) radiation, (d) synchronized time trace of Idiv. The area highlighted in gray in figure (a) indicates the time window when the Li-beam is switched off to enable the background measurements. The inter- and intra-ELM phase are highlighted in figure (d) in blue and red, respectively. correct for the so-called “shine-through” effect at the plasma edge [19]. This diagnostic setup was already used in the investigation of the neand Terecovery after the ELM crash [7]. In this work, a new approach to determine the electron density profile from Li- BES during the ELM has been developed by consistently correcting the measured signals from the enhanced passive radiation observed during the intra-ELM phase. Figure 1a shows the measurements of one Li-BES channel during an ELMy H-mode where the gray shaded area indicates when the beam is switched off in order to measure the background radiation. The signal is compared to the divertor shunt current Idiv (Fig. 1b) showing that both the active and passive radiation are strongly modulated by ELMs (see dashed vertical lines). To properly calculate intra-ELM density profiles, an algorithm for the background correction was developed in which the background radiation is synchronized relative to the ELM onset using Idiv as a reference signal. This is then subtracted from the synchronized signal when the Li-beam is not switched off in order to obtain only the active part. In Figure 1c the conditional synchronized background signal (blue) and the sum of active and passive signals (red) are shown. The synchronized divertor shunt current is instead shown in figure 1d where the intra- and inter-ELM phases are highlighted in red and blue, respectively. This analysis is applied to every line of sight (LOS) and hence intra-ELM density profiles can be obtained through the IDA forward model. Note that the intra-ELM electron density measurements are also possible with Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 4 high-frequency chopping [20], however, this mode of operation is not available anymore. 2.2. Ion and flow measurements In fusion plasmas, the ion temperature, the plasma poloidal and toroidal flows are typically measured via charge exchange recombination spectroscopy (CXRS). The ASDEX Upgrade tokamak is equipped with several CXRS systems which cover the plasma, from the low to the high field side [21, 22, 23]. A recent upgrade of the edge CXRS diagnostic provides a full reconstruction of the impurity poloidal and toroidal flows, density, temperature, and hence Erprofiles at a minimum time resolution of 10 µs [24]. Note that Eris calculated from the radial force balance equation (RFB) Er=∇r(nαTα)/(eZαnα)+vφ,αBθ−vθ,αBφby measuring every term for a specific plasma species α, in this work helium ions. While the E×Bvelocity is a property of the plasma, the single contributions in the RFB equation are different for every αand hence, for example, the poloidal and toroidal velocity of the main ions differ from the helium ones (more details con be found in [25]). High frequency measurements of Ercan also be obtained via Doppler reflectometry. However, these measurements are often not possible in H-mode because the signal to noise ratio is low due to the low turbulence level at the edge. In the cases discussed here, the He2+(n= 4 →3) transition at 468.571 nm is used for which nine LOS can be measured simultaneously. Five are used for measuring the poloidal projection of the He2+ velocity and four for the toroidal one. Conventional line fitting of passive and active radiation provides ion temperature and flow velocity measurements from the Doppler width and Doppler shift, while the impurity density is derived from the radiance of the emission line. In this work, particular effort has been devoted to the flux surface mapping of the CXRS measurements. A pressure constrained equilibrium reconstruction coupled with the flux diffusion equation has been used to map the different channels of all diagnostics including the CXRS system every 1 ms [26], while the reconstruction of the outer mid-plane separatrix position, available every 100 µs, has been employed to correct the mapping at the appropriate time resolution. To do that, the pressure constrained equilibrium has been rigidly shifted at every measurement time point relatively to the outer mid-plane separatrix position. Note that the latter is derived from an equilibrium reconstruction based only on magnetic measurements . In this way, the CXRS profiles can be mapped also during the intra-ELM phase where the plasma is usually observed to move inward by roughly 0.5 to 1.0 cm which is more than the diagnostic radial resolution, which is between 0.3 and 0.5 cm. Note that this procedure is mostly uncertain in the time window right after the ELM crash, i.e. in this case within 2 ms after the ELM crash, where the plasma shows the strongest inward movement. Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 5 Figure 2: Timetraces of one representative discharge: (a) neutral beam input power, (b) line averaged edge (blue) and core (red) electron densities, (c) radial excursion of the plasma at the outer midplane, (d) divertor shunt current. The time window highlighted in gray has been excluded from the analysis due to the temporary drop of the heating power. 3. Edge profile evolution through-out the entire ELM-cycle Several dedicated plasma discharges were performed to obtain high resolution measurements during the ELM cycle. In total 80 ELMs were selected to apply the analysis methods as described in Section 2 of which roughly 1/3 are used for the background measurements of the Li-BES diagnostic. The fast edge CXRS also works in a burst mode and it acquires only 2/3 of the total discharge. Hence the overall number of ELMs used in the synchronization is similar between the different diagnostics. The experiments are type-I ELMy H-mode plasmas with a toroidal magnetic field on-axis of −2.5 T, a plasma current of 1 MA, a core line-averaged density of approximately 7×1019 m−3and a total heating power of 8.7 MW. Figure 2 shows time traces of the neutral beam input power (a), the line averaged density at the edge (blue) and in the core (red), the outer mid-plane plasma excursion (c) and the divertor shunt current (d). The time window highlighted in gray has been excluded from the analysis due to the temporary drop of the heating power which may affect the plasma conditions. The plasma is radially scanned through the views of the LOS to improve the radial coverage of the measurements while the other plasma parameters are kept constant. A rather constant ELM behaviour is observed in this time window with a frequency of around 85 Hz. Note that every ELM perturbs the position of the plasma by approximately 5 to 6 mm (see figure 2c). This has been taken into account when mapping the CXRS Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 6 Figure 3: (a): Contour plot of the synchronized ion temperature profile as measured during an ELM cycle. (b): The divertor thermal current used as the reference signal for the synchronization. measurements onto flux coordinates as described in Section 2.2. The divertor shunt current Idiv has been used as ELM monitor signal and consequently as the reference for the conditional average of the measurements. Figure 3a shows the contour plot of the synchronized ion temperature profiles where the colors indicate the measured temperatures. The synchronized reference signal, i.e. Idiv, is shown in figure 3b. After t−tELM = 0 s, i.e. the ELM onset, an increase of the temperature close to the separatrix is observed while the temperature at the pedestal drops. The length of this phase, usually referred to as the ELM-crash (or intra-ELM phase), is roughly 4 ms long and in general depends on the specific plasma conditions [27]. Shortly after the ELM onset, a strong reduction of Tiis observed similar to observations in DIII-D [8]. To characterize the evolution of the edge Ti,nHe2+ profiles and their gradients, the synchronized profiles have been fitted with a spline function. The resulting Tiand nHe2+ profiles from the spline fit were cross-checked with binned data showing good agreement within the uncertainties. The advantage of the spline fit in comparison with the direct 2D (time and space) binning of the data is the determination of the gradients. The splines are C2piecewise-polynomial functions that constrain the derivative to be continuous and derivable [28]. This is not possible if the gradients are calculated simply as differences between the binning points, which then results in an unphysical scatter. For the determination of Er, the spline fits are only necessary when evaluating Tiand nHe2+ , while for vHe2+ θand vHe2+ φstandard 2D binning can be used since their gradients are not required (see radial force balance). The IDA algorithm is also based on spline functions for the reconstructed profiles thus providing the gradients automatically. This also allows for the application of 2D binning of neand Temeasurements and their Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 7 Figure 4: Evolution of the edge profiles relative to the ELM onset through the entire ELM cycle: (a) ion temperature, (b) electron temperature, (c) He2+ density, (d) electron density, (e) He2+ poloidal velocity, (f) He2+ toroidal velocity. The four different colors indicate the different radial positions, reported in ρpol. gradients. To avoid unnecessary smoothing, a time resolution of 50 µs for the binning and the fitting has been chosen, which is shorter than the acquisition time of the system, i.e. 65 µs for the CX measurements and 100 µs for the IDA algorithm. This approach was chosen in order to investigate the statistical scatter of the measurements without having to evaluate the uncertainties of the spline fits and of the profile gradients, which can not be unambiguously determined. Figure 4 shows the evolution of the edge profiles throughout the entire ELM cycle. The profiles are reported as time-traces at four different radial locations (see colors) within the outermost 1.5 to 2 cm. The figures 4a and 4b show the ion and the electron Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 8 Figure 5: Edge profiles at three distinct time windows (as highlighted in figure 4) relative to the ELM onset: (a) ion temperature, (b) electron temperature from Thomson scattering (red) and from the IDA algorithm (black), (c) He2+ density, (d) electron density from Thomson scattering (red) and from the IDA algorithm (black), (e) He2+ poloidal velocity, (f) He2+ toroidal velocity. The three different color panels indicate the different time windows. temperature, respectively, while in figures 4c and 4d the He2+ density and the electron density are presented. Additionally, the helium toroidal and poloidal flows are shown in figures 4e and 4f. The time traces are shown relative to the ELM onset defined by the rise of the divertor shunt current (see figure 3b). Note that the ELM takes place at the mid-plane and the expelled heat and particle fluxes reach the divertor few hundred of micro-seconds later. Therefore the synchronized profiles are affected from the ELM at slightly negative times. The complete edge profiles measured during the highlighted time windows are instead reported in figure 5 in which the electron density and temperature from the IDA algorithm are compared to the Thomson scattering data (fig. 5b and 5d). Pedestal and Erprofile evolution during an ELM cycle at ASDEX Upgrade 9 The behaviour of Tiat the ELM onset is found to differ substantially from that of the electron temperature. The heat pulse originating in the steep gradient at ρ≈0.98 leaves the electron temperature close to the separatrix almost unperturbed while Ti is observed to increase (compare blue and green time traces) in agreement with the results from DIII-D [29]. This might be due to the combination of two effects: (i) the electron cyclotron emission close to the separatrix is dominated by the so called “shine through” which complicates the reconstruction of Teprofiles, (ii) in order to recover the measurements in that region a forward modelling of the ECE emission has been applied in which a two-point model forces the separatrix Teto be close to 100 eV. This condition is based on physics assumptions of a high parallel electron heat conductivity in the SOL and thus should also hold during the ELM crash [30]. To test the hypothesis, Thomson scattering measurements have been evaluated and, within the error bars, no detectable changes of the separatrix Teduring an ELM cycle have been observed at the very edge of the plasma (see figure 5b), as reported in [29]. Further inside the plasma, towards the pedestal top region, Tiand Teshow a similar behaviour (black and red time traces in figures 4a,b). The phase of enhanced Tinicely correlates with the detection of hot filaments (Tiup to 100 eV) during the ELM-crash [31]. The electron density during the ELM crash shown in figure 4d has been evaluated using the conditional synchronization approach described in section 2.1. The evolutions of neand nHe2+ throughout the entire ELM cycle are similar: after the flattening of the gradient, the pre-ELM profiles are reestablished on a shorter time scale compared to Te, as already reported in [7]. However, during the first 2 ms after the ELM onset a stronger increase of nHe2+ compared to neclose to the separatrix is observed (figures 4c and 4d). This difference is probably connected to the different charges of the considered ions. A higher temperature (energy) is required to fully ionise helium compared to deuterium. Hence, the nHe2+ profile is shifted inwards compared to the neprofile and consequently also the pivot point during the ELM crash. Therefore a much stronger increase of the nHe2+ profile close to the separatrix is expected. Note that a small increase of neat the separatrix is also observed (figure 5d) at the ELM crash which propagates outwards in the SOL indicating that the pivot point is close to the separatrix as reported in previous publications [32, 33, 20]. This is consistent with the observations in DIII-D [29] in which the pivot point of the C6+ density profile is observed to be further in due to the higher ionisation energy. The behaviour of the poloidal and toroidal velocities appear more puzzling (figures 4e,f): the gradient of vHe2+ θshows a collapse at the ELM onset similar to the other profiles but, just after it, a fast poloidal spin-up is observed and the pre-ELM profile is re-established by its slowing down. Note that these discharges were performed in deuterium with roughly 12% helium. Therefore, the behaviour of vHe2+ θis specific only to He2+ as an impurity ion and is different from the main ion poloidal rotation or any other impurities. For instance, this effect has not been observed in the recovery of the N7+ poloidal velocity. The helium toroidal velocity represents a reversed shear moving towards the separatrix, as reported in [34]. At the ELM crash, the shear suddenly collapses to a flat profile and shortly after, the whole edge profile drops. Later, vHe2+ φ