3D simulations of gas puff effects on edge plasma and ICRF coupling in JET

PAPER 3D simulations of gas puff effects on edge plasma and ICRF coupling in JET To cite this article: W. Zhang et al 2017 Nucl. Fusion 57 056042 View the article online for updates and enhancements. Related content Effects of outer top gas injection on ICRF coupling in ASDEX Upgrade: towards modelling of ITER gas injection - 3D simulations of gas puff effects on edge density and ICRF coupling in ASDEX Upgrade - Maximization of ICRF power by SOL density tailoring with local gas injection - Recent citations The modeling of atom – neutral collisions for beam emission spectroscopy applications O. Asztalos et al - Impact of ICRF on the scrape-off layer and on plasma wall interactions: From present experiments to fusion reactor V. Bobkov et al - Optimization of discharges with ion cyclotron range of frequencies using local gas injection in EAST G. Urbanczyk et al - This content was downloaded from IP address 87.218.223.151 on 10/08/2020 at 17:18 1 © 2017 Max-Planck-Institut fur Plasmaphysik Printed in the UK 1. Introduction Plasma heating with electromagnetic waves in the ion cyclotron range of frequencies (ICRF) is one of the most promising auxiliary plasma heating methods in magnetic controlled fusion devices. It has been successfully implemented in present-day tokamaks and stellarators and will be applied in future fusion machines such as ITER and DEMO. As the biggest tokamak in operation at present, joint European torus (JET) is equipped with a robust ICRF heating system—four 4-strap A2 ICRF antennas (powered per pair) and one ITER-like-antenna (ILA)—capable of providing a total radio frequency power of ~8.0 MW in ELMy H-Mode plasmas [1]. For Nuclear Fusion 3D simulations of gas puff effects on edge plasma and ICRF coupling in JET W.Zhang1,2,3, P.Jacquet4, E.Lerche4,5, R.Bilato2, V.Bobkov2, D.Coster2, Y.Feng2, C.Guillemaut6, M.Goniche7, D.Harting4, T.Lunt2, J.-M.Noterdaeme1,2, G.Szepesi4, D.Van Eester5 and the JET Contributors8,a 1 Applied Physics Department, University of Ghent, Ghent, Belgium 2 Max-Planck-Institut für Plasmaphysik, Garching/Greifswald, Germany 3 Institute of Plasma Physics, Chinese Academy of Sciences, Hefei, People’s Republic of China 4 CCFE, Culham Science Centre, Abingdon, OX14 3DB, United Kingdom 5 LPP-ERM/KMS, EUROfusion Consortium Member, Trilateral Euregio Cluster, Brussels, Belgium 6 Instituto de Plasmas e Fusão Nuclear, Instituto Superior Técnico, Universidade de Lisboa, Lisboa, Portugal 7 CEA, IRFM, F-13108 Saint-Paul-Lez-Durance, France 8 EUROfusion Consortium, JET, Culham Science Centre, Abingdon, OX14 3DB, United Kingdom E-mail: [email protected] Received 13 September 2016, revised 24 February 2017 Accepted for publication 21 March 2017 Published 7 April 2017 Abstract Recent JET (ITER-Like Wall) experiments have shown that the fueling gas puffed from different locations of the vessel can result in different scrape-off layer (SOL) density profiles and therefore different radio frequency (RF) coupling. To reproduce the experimental observations, to understand the associated physics and to optimize the gas puff methods, we have carried out three-dimensional (3D) simulations with the EMC3-EIRENE code in JET-ILW including a realistic description of the vessel geometry and the gas injection modules (GIMs) configuration. Various gas puffing methods have been investigated, in which the location of gas fueling is the only variable parameter. The simulation results are in quantitative agreement with the experimental measurements. They confirm that compared to divertor gas fueling, mid-plane gas puffing increases the SOL density most significantly but locally, while top gas puffing increases it uniformly in toroidal direction but to a lower degree. Moreover, the present analysis corroborates the experimental findings that combined gas puff scenarios—based on distributed main chamber gas puffing—can be effective in increasing the RF coupling for multiple antennas simultaneously. The results indicate that the spreading of the gas, the local ionization and the transport of the ionized gas along the magnetic field lines connecting the local gas cloud in front of the GIMs to the antennas are responsible for the enhanced SOL density and thus the larger RF coupling. Keywords: gas puffing, ICRF coupling, 3D simulation, edge plasma, JET (Some figuresmay appear in colour only in the online journal) W. Zhang etal Printed in the UK 056042 NUFUAU © 2017 Max-Planck-Institut fur Plasmaphysik 57 Nucl. Fusion NF 10.1088/1741-4326/aa6817 Paper 5 Nuclear Fusion IOP International Atomic Energy Agency a See the author list of ‘Overview of the JET results in support to ITER’ by X. Litaudon etal to be published in Nuclear Fusion Special issue: overview and summary reports from the 26th Fusion Energy Conference (Kyoto, Japan, 17–22 October 2016). 2017 1741-4326 1741-4326/17/056042+15$33.00 https://doi.org/10.1088/1741-4326/aa6817 Nucl. Fusion 57 (2017) 056042 (15pp) W. Zhang etal 2 ITER, the ICRF heating system being developed is designed to provide a total heating power of 20 MW [2]. ICRF heating relies on the fast wave to transfer the radio frequency (RF) power launched from the antennas into the plasma. However, the ICRF power coupling depends critically on the width of the evanescent layer Revan in front of the launchers, namely on the distance between the fast wave cut-off density layer and the antenna straps [3]. The amount of ICRF power coupled to the plasma can be expressed as ∝ PVRZ 2 coupled max 2 c c 2 / , where Vmax is the anti-node voltage in the transmission line which is usually limited by arcing in it, Zc is the characteristic impedance of the transmission line. Given Vmax and Zc values, P coupled is only proportional to the coupling resistance Rc. However, the coupling resistance decreases exponentially with Revan , i.e. ∝α−⋅ R R eR c0 evan , where α is a tunneling factor that mainly depends on the antenna properties (e.g. phasing) [4]. Thus, to ensure the best performances of the ICRF heating system, it is crucial to minimize this evanescent distance by increasing the plasma density in front of the antennas to a level above the fast wave cut-off density. In JET H-mode plasmas the fast wave cut-off density is typically ×2.0 1018 m−3 for dipole phasing operation [5]. The local scrape-off Layer (SOL) density can be tailored by changing the gas puff locations and the gas puff rate. Previous experiments and simulations indicate that the SOL density can be increased with main chamber (top or mid-plane) gas puffing instead of divertor gas puffing [5–10]. A pertinent question related to the use of local gas to improve ICRF wave coupling is the impact of this technique on the plasma param eters. This was discussed in some detail in [5], and repeated here for completeness. The measured plasma core kinetic profiles (Te and ne) were not modified whether divertor, top or OMP gas was used. This is illustrated in figure1, where the time trace of the electron temperature at the plasma centre, and electron plasma energy are plotted for pulses with divertor and OMP gas injection respectively. The electron temperature and electron density profiles for these two pulses are plotted in figure1(e). Profiles are taken before the sawtooth crashes; data are averaged over five time windows of 0.1 s each as shown in figure1(c). In these plasma conditions, fast wave heating results predominantly to an electron heating via slowing down of the fast H ions, with a typical centre electron temperature response of ~0.5 keV MW−1. The electron kinetic profiles overlay perfectly, suggesting that within measurement errors bars, the fast wave propagation and absorption characteristics as well as the transport properties are the same, independent of the gas injection location. An important issue when considering gas injection from different locations is the consequence of using main chamber instead of divertor gas on plasma energy confinement: In JET no degradation of pedestal pressure (figure 1(d )) and confinement was observed when using the mid-plane or top inlets instead of the divertor inlets at the same gas fuelling level [5]. Another illustration is show in figure2 where, for the series of pulse studied in this paper, the plasma energy if plotted as a function of the injection rate for different gas valve locations, showing that plasma energy was not affected by gas location. Note that in this experiment the main purpose was to measure the antenna coupling resistance reliably: the ICRH power was constant from pulse to pulse (launched-power was a set-point in a ICRF system control loop) and the power level was modest (~3 MW); hence increase of antenna coupling resist ance when using local gas puffing did not lead to higher ICRF power into the plasma, but rather it reduced the voltage at the antennas. The quantity of gas required for antenna coupling improvement is also a key factor, particularly since increasing gas dosing can have a negative impact on plasma energy confinement in existing tokamaks [11]. In this context it is important to emphasize that the antenna coupling improvements reported here can be obtained when using moderate gas rates: in practice, the gas puff rate used in scenario development aiming at best D-T target plasmas is between 0.5 × 1022 electronss−1 (hybrid) and 2 × 1022 electrons s−1 (baseline) [12], while the improvement in the antenna loading in dedicated ICRF coupling experiments was ~50% when fueling the plasma Figure 1. JET time traces and electron kinetic profiles when using divertor or OMP gas injection in JET. The time traces are (a) the auxiliary heating power, (b) the D2 gas injection rate, (c) the electron temperature at the plasma centre, and (d ) the total electron energy. Details of the Te (left y-axis) and ne (right y-axis) profiles are shown in (e). The profiles are averaged over the time intervals indicated in grey on the time trace graphs. Error bars correspond to the scattering in the data. Reproduced courtesy of IAEA. Figure from [5]. Copyright 2016 IAEA. Nucl. Fusion 57 (2017) 056042 W. Zhang etal 3 only with nearby mid-plane inlets instead of divertor inlets at an injection rate of 1.0 × 1022 electrons s−1 [5]. Hence the local gas injection technique to optimize ICRF heating performance is now routinely used in JET in the development of plasma scenarios. The success of experiments and simulations kindled the interest in enhancing ICRF coupling via local gas puffing in many tokamaks, such as ASDEX Upgrade (AUG) [7–9], DIII-D [13, 14], EAST, JET [5, 10], KSTAR, WEST [15], ITER [16, 17] and DEMO. For ITER in particular, the width of the evanescent layer Revan is larger than the one in present and past machines [10]. Moreover, the SOL density has large uncertainties in different plasma scenarios, making it difficult to make accurate predictions of the total RF coupled power. It is thus extremely important to investigate how to tailor the SOL density in front of the ICRF antennas through the local gas puffing methods and to have a suite of numerical tools validated against present experiments. This paper mainly concentrates on the numerical studies of deuterium gas injection from different locations in JET-ILW while various experimental data have been used for comparisons. Several aspects were considered to be important in describing accurately the gas puff experiments using numer ical simulations. First of all, the simulations have to be 3D since local gas puffing induces toroidal and poloidal inhomogeneous densities in SOL. Second, the neutrals transport and ionization by electron impact and plasma transport have to be computed accurately in realistic geometries. Third, the main chamber plasma facing components have to be taken into account because they play an important role in generating the recycling fluxes and acting as boundary conditions. Fourth, the gas valves and gas pumps need to be described realistically. EMC3-EIRENE [18] is a 3D edge plasma fluid and neutral transport code which fulfills all the requirements mentioned above. The paper is dedicated to the applications of the EMC3-EIRENE simulations in JET. To calculate the coupling resistances in different gas puffing scenarios, simulations via combining the EMC3-EIRENE simulations and the antenna code simulations are necessary. This only requires the import of the 3D density and temperature profiles from the EMC3-EIRENE simulations into the antenna codes. In addition, this work is part of a continuous effort to valid ate the EMC3-EIRENE code using available data from present tokamaks in order to carry out future simulations for ITER with confidence and to define best gas injection locations to optim ize ICRF antenna coupling. This work completes the already published simulations for AUG [8]. AUG is a middle sized tokamak with major radius/minor radius = 1.65m/0.5– 0.8 m, plasma volume = 13 m3 and BT/Ip = 2.5 T/0.8 MA. JET is a larger tokamak with major radius/minor radius = 2.96 m/1.25–2.10 m, plasma volume = 100 m3 and BT/Ip = 2.7 T/2.5 MA. A full-tungsten wall is used in AUG while an ITER-like wall (tungsten divertor and Beryllium main chamber wall) is used In JET. Both in these two machines, the divertor, the mid-plane and the top gas puffing methods are investigated. The gas valve settings are different in the two machines. For instance, in AUG the top gas is usually puffed from a gas valve located in the inner top of the vessel and the resulting gas is rather localized. However, in JET the top gas is puffed right from the top of the vessel and the generated gas is poloidal widely spread. Compared to the previous work on AUG [8], the following differences/advancements are made in the paper: (1) inclusion of a divertor gas pump in the simulation model; (2) more elaborated settings of the top and midplane gas valves; (3) comparisons of the experimental and simulated neutral pressures; (4) studying the influences of the transport parameters on the ICRF coupling; (5) investigations of the combined gas puffing. Note that density modifications induced by the ICRF induced E × B drifts [19, 20] and ponderomotive expulsion [21] are not included in the EMC3-EIRENE code. Moreover, no ionization induced by ICRF wave [22, 23] is considered in our simulations. In fact, evidences of RF ionization are only found when the antenna power is lower than 20 kW in AUG [24]. It is worth mentioning that local gas puffing can, in addition to improving the antenna coupling, also decrease the impurity influxes from antennas [7]. Changing the local density can also potentially modify the heat loads on the antenna structure, as attest scaling laws from JET experiments both with C-wall [25] and ITER-like wall [26], where it was found experimentally that heat fluxes increase with the local density in front of the antennas and the RF voltage in the transmission lines feeding the antennas. It is however observed in practice that local gas injection is also beneficial on JET to reduce the heat loads at the limiters close to the antennas. The reason might be of two folds: (a) The density increase in front of the antenna and coupling improvement reduces the antenna parallel electric field that drives RF sheaths rectification and RF specific heat loads. (b) Gas injection could also locally reduce 0.5 1 1.5 2 x 10 22 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 gas rate (e/s) W mhd (MJ) DIV MID TOP MID+TOP Figure 2. Plasma energy evaluated from EFIT for the series of pulses with divertor, mid-plane or top gas analysed in this paper. Note most of the Wmhd values for top puffing (6 out of the 8) were generated at slightly lower plasma current of Ip = 2.4 MA instead of 2.5 MA. Nucl. Fusion 57 (2017) 056042 W. Zhang etal 4 the electron temperature which again is beneficial from the point of view of RF specific heat loads and impurity release. Besides local gas puffing, reducing the radial distance from the separatrix to the ICRF antennas is also a potential way to increase the ICRF coupling. However, this technique may significantly increase the power fluxes to the main chamber limiters (including the antenna limiters) or blanket module (in ITER). Also the minimum gaps in ITER will be constraint by the response time of the plasma control system. In comparison, gas puffing is a more flexible and efficient way to improve the ICRF coupling. Moreover, local gas puffing is able to reduce the impurity influxes from antennas, as discussed before. Despite these merits, local gas puffing can influence the diagnostics nearby. The paper is organized as follows. In section2, the simulation setup is presented, including the settings of the gas puff and gas pump, the plasma facing components, the computational grid and the validations of the background plasmas. In section3, the simulation results are discussed. These include the neutral and electron densities during different gas puffing, the calculations of the ICRF coupling resistances, the mechanisms of gas puff effects on SOL density and ICRF coupling. In section4, the combined gas puff scenarios are investigated. In section5, conclusions and an outlook are given. 2. Simulation setup EMC3-EIRENE is a code package which self-consistently couples the Edge Monte-Carlo 3D plasma fluid code (EMC3) [18] and the kinetic neutral Monte Carlo code (EIRENE) [27]. EMC3 solves a set of time-independent Braginskiilike equations for mass, parallel momentum, electron and ion energy with anomalous transport coefficients. EIRENE is a Monte-Carlo solver of the Boltzmann equationfor the neutrals. The parallel transport is treated purely neoclassically in EMC3 while the perpendicular transport coefficients ⊥ D and χ⊥ are specified as free parameters. These free parameters can be derived through comparisons of the simulated plasma parameters with the experimental ones. Besides its wide applications in stellarators [28, 29], EMC3-EIRENE is also attracting extensive interest for its implementation in tokamaks where some 3D SOL physics aspects need to be accounted. For example, the code was applied to model ICRF antenna coupling experiments [6, 8], the influences of the Resonant Magnetic Perturbation fields [30], the 3D edge plasma convection [20] and the plasma–wall interactions [20, 31]. Note that drifts and volume recombination are not calculated in the present versions of the EMC3-EIRENE code. It is worth mentioning that progresses have been made in applying prescribed drifts in the code [20]. These prescribed drifts can be any drifts while they have to be calculated in advance and read as input fields. Before running our simulations, it is important to first build a realistic simulation model. The layout of the JET torus is illustrated in figure3, including the main wall structures, the gas injection modules (GIMs) and the locations of the key density measurements. A toroidal 360° computation grid is built based on the EFIT equilibrium [32] from discharge #84476 at 50.0 s (see figures4 and 5). Toroidally, the computational grid is composed of sixteen equally constructed segments and each segment is divided into sixteen parts. The JET vacuum vessel is divided toroidally into eight octants, thus two segments in our simulations are used to represent one octant. A further increase of the toroidal resolution of the grid results in a noticeable increase of the Monte Carlo noise (both in EMC3 and EIRENE). In the following, 0° is set at the boundaries of octant one and octant eight. In the poloidal cross-section, the grid is divided into the core, SOL and private flux region Figure 3. (a) Top view of JET illustrating the toroidal locations of the plasma facing wall components, the GIMs and the key density diagnostics. Adapted with permission from [35]. © 2014 Euratom. (b) Poloidal cross-section of JET shows the gas route puffed from different GIMs. Reproduced courtesy of IAEA. Figure from [5]. Copyright 2016 IAEA. Nucl. Fusion 57 (2017) 056042 W. Zhang etal 5 (PFR). In the SOL the grid resolution is 39 × 481 × 256 (radial × poloidal × toroidal) cells. A relative high resolution is specified in the regions with physics of interest: the divertor and the outer mid-plane regions in front of the limiters. In addition to the EMC3 grid, the EIRENE grid includes an ‘additional region’ (figure 4, in yellow color) bounded by the vessel wall so that the neutral transport inside the whole vessel can be calculated. A toroidally symmetric gas pump (cryo pump) is set at its actual position. This pump is the dominant gas pump and the contributions of other gas pumps (such as turbo pump) are negligible for the experiments investigated. The reflection coefficient of the pump surface is set as 0.98, meaning that 2% of the neutrals hitting the pumping surface are absorbed while the rest is reflected. With this setting, the gas pump rate is equal to the gas puff rate (1.7 × 1022 electrons s−1) and the electron density matches the experimental one in our simulations. Two neutral baffles (the purple dashed lines in figure4) are set in regions outside the divertor. This is to make the neutral transport in those regions as realistic as possible. In reality hardly any gas can pass through the locations of the neutral baffles. In steady state JET H-mode plasmas, the recycling fluxes from the divertor and main chamber wall are the primary particle sources. In particular, the limiters in the low field side have strong interactions with the plasma. They play a crucial role in confining the plasma and generating recycling fluxes. Thus, all these limiters including eight Wide Poloidal Limiters (WPLs), two ITER-Like-Antenna (ILA) limiters and one Narrow Poloidal Limiter (NPL) are taken into account in our simulations (see figure5). The different types of limiters have different widths and different radial positions. As a result they will induce toroidal inhomogeneities in the SOL. This will be discussed later in more detail. The toroidal axisymmetric plasma facing components in our model include the inner and outer targets and the roof baffle. Because the inner limiters (those on the high field side) are identical and axisymmetric, the inner wall is placed at the same location as the leading edges of these inner limiters. For the wall materials, tungsten (W) and Beryllium (Be) corresponding to the divertor and main chamber wall respectively, are used in our simulations. No sputtering is included in the simulations, the use of different wall materials affects the reflection model in EIRENE. In our simulations, the GIMs are either set at the realistic positions or set in such a way that the simulated gas is equivalent to the experimental one (in terms of the spreading and intensity of the gas). For the top gas puff (GIM7 or GIM8), eight poloidally evenly distributed point sources in the top of the vessel are used to inject the same amount of gas as in the experiment. We have used eight point sources so that the poloidal distribution of the gas is homogeneous enough. For the divertor gas puff (GIM11), eight toroidally evenly distributed point sources have been set at the realistic divertor positions to generate the experimental equivalent gas. The settings of the mid-plane GIMs (GIM3, GIM4 or GIM6) are more complicated. This is because the midplane GIMs are located very deep in the so called A-port. They are far away from the main chamber and are outside our simulation domain. As the gas is transported radially inward from the GIM, it keeps spreading poloidally and toroidally filling essentially the whole A-port. A rather homogenous gas is found at the cross-section where the A-port is connected to the main chamber. To simulate this inside the vessel, we have put five point sources distributed equidistantly in poloidal direction (~50 cm away from the seperatrix) at the outer mid-plane to simulate the experimental gas puff as realistically as possible. Before the gas reaches the hot plasma and is ionized, it has already spread Figure 4. A poloidal cross-section of the EMC3-EIRENE computation grid. Figure 5. 3D structure of the computational grid, the limiters and the antennas. Nucl. Fusion 57 (2017) 056042 W. Zhang etal 6 to a rather large extent. At the leading edge of the limiters, the spreading of the gas is in the range of (−0.85 m, 0.85 m) poloidally and about ten degrees toroidally (depending on the wall structures nearby). These features are well captured in our simulations. The setup procedure of the JET mid-plane GIMs in our simulations is similar to the one implemented in AUG (see [8]) but more sophisticated. In AUG only one mid-plane GIM is needed to generate the experimental equivalent gas. The simulation strategy is as follows. First, a reference pulse with symmetrical divertor gas puffing is simulated, and in particular transport parameters are adjusted to match the measured profiles of ne, Te and divertor saturation current. This validated plasma is then assumed as background plasma, and it is used in the simulations to study the changes in the SOL when puffing gas from other locations, for example at specific toroidal locations at the outer mid-plane or at the top of the vessel. Our divertor reference is the JET pulse 84476 where the divertor GIM11 was used. The separatrix averaged density is2 × 1019 m−3 and the net input power is 14 MW, Pnet = PNBI + PICRH + POH − Prad. These two parameters are derived from experimental measurement and they are regarded as inner boundary conditions. Note that experimentally these parameters remained unchanged when further using top or mid-plane gas puff. Because no impurities are considered in our simulations, the power radiated by impurities has been subtracted from the total heating power to evaluate the net power. For the discharges considered in our studies, BT = 2.7T, Ip = 2.5 MA, Prad = 6 MW and the minimum distance from the separatrix to the wall is 6 cm. To correctly describe the cross-field transport of the plasma, the upstream and downstream profiles in our simulations are best matched to experiments by adjusting the perpendicular plasma transport parameters (i.e. the so-called ‘free parameters’ mentioned above). The results of this procedure are shown in figure6. The upstream profiles Figure 6. Matching the background plasma parameters with experiments: (a) perpendicular plasma transport coefficients ⊥ D and χ⊥ ; (b) comparisons of mid-plane electron density, the error bars correspond to the statistical dispersion of the reflectometry data; (c) comparisons of mid-plane electron temperature; (d ) comparisons of particle fluxes to the divertor (around the outer strike point in the roof baffle). Nucl. Fusion 57 (2017) 056042 W. Zhang etal 7 include the mid-plane electron density ne and the mid-plane electron temper ature Te. The downstream profiles include the particle flux to the divertor targets jsat (the outer strike point is located on the right bottom of the divertor). The perpendicular transport parameters include the particle and heat diffusion coefficients ⊥ D and χ⊥ , in which χχχ== ⊥⊥⊥ie . For the experimental data, we have selected the values during the full inter-ELM periods. The averaged mid-plane ne is obtained with the following procedures: the core and pedestal density are derived from High resolution Thomson Scattering (HRTS), and the SOL profiles are derived from the reflectometer (more than 2700 reflectometer profiles are used in this averaging procedure). The density data are averaged over the time 49.5–50.5 s. The mid-plane Te is averaged from the HRTS data and is only reliable in the plasma core and in the near SOL. In the far SOL, the measured Te with HRTS has large error bars. The particle fluxes to the divertor targets jsat are measured with Langmuir probes. jsat is equal to the current (i.e. the particle stream) in a flux tube divided by the area perpendicular to that flux tube. 3. Simulation results Once the background plasma is defined, the gas source is switched from the divertor to the main chamber (top or midplane) of the machine in our simulations. All other parameters, including the total gas puff rate 1.7 × 1022 electrons s−1, are kept unchanged. The JET pulses studies using this procedure are listed in table1. 3.1. Gas puff effects on the edge plasma After running the simulations for all the cases, the simulated neutral and electron densities during gas puffing in the divertor, top and mid-plane are shown in figure7. nD2, nD and ne represent the deuterium molecular density, the deuterium atom density and the electron density, respectively. In each case, the poloidal cross-section is taken at the toroidal position where the GIM is located. According to our simulation results (figure 7), a neutral density cloud localized in front of the top or mid-plane GIMs develops during the corresponding gas puff. As the gas cloud interacts with the plasma, it is ionized locally when the electron temperature is larger than the threshold ionization temperature (~2 eV). The neutral ionization rate depends nonlinearly on the electron temperature [33]. The complex ionization processes are calculated in EIRENE based on the AMJUEL/HYDHEL database and an electron density cloud is formed in front of the GIMs where the ionization happens. The cloud of enhanced density can be seen in figure7 (right column). The extension of the density cloud mostly depends on the spreading of the neutral gas while the peak density largely depends on the ionization rate. In the density cloud, the electron temperature is lower than the background plasma temperature. Due of the rapid motion of charged particles along field lines (i.e. the parallel transport), density in regions which are magnetically connected to the density cloud are soon enhanced. The density increase in the SOL depends on the structures of the field lines, on the density cloud as well as on the perpendicular transport. If these field lines spread widely, the high density source is shared by a large region so that the increase of density is global but not substantial; if the field lines remain close to each other over a long distance, the density increase can be significant but more local. Comparisons of the simulated gas pressure and the exper imental one are made for the mid-plane gas puff case (GIM4) (figure 8). It is shown that the maximum value of the gas pres sure (gas pressure at the toroidal position of the GIM) and the toroidal distributions of the gas pressure are well reproduced in our simulations. The local gas pressure increases (the small peaks) are due to the local recycling at the limiters. The quanti tative agreements of the gas pressure confirm that the settings of the mid-plane GIMs in our simulations are reasonable. Furthermore, we have compared the experimental and simulated electron density for the divertor (GIM11) and mid-plane (GIM4) gas puff cases (figure 9). The compariso ns are made at the reflectometer position, i.e. toroidally at φ = −52.5° and vertically at the plasma outer mid-plane. Only the reflectometer data during the inter-ELMs phases is used. Quantitative agreement is found between the simulated and experimental density profiles. A match within 95% is seen during divertor gas puffing. Small differences in the limiter shadow and in the near SOL are observed during mid-plane gas puffing. Nevertheless these differences are still within the experimental error bars. This good agreement further confirms that Table 1. The simulated cases in parallel with the experiments. Simulated cases Gas valve positions (toroidal) Corresponding experiments Gas puff rate Divertor gas puff GIM11 Toroidally evenly distributed #84476, #84485, #90140 1.7 × 1022 electrons s−1 Top gas puff GIM7 φ = −137.5°#84739 GIM8 φ = −45°#84740 Mid-plane gas puff GIM3 φ = −135°#90131, #90133, #90136, #90137 GIM4 φ = 0°#84478, #90128, #90129, #90130 GIM6 φ = 108°#84477 Nucl. Fusion 57 (2017) 056042 W. Zhang etal 8 the settings of the mid-plane GIMs in our simulations are reasonable, and gives us confidence that our simulations are able to reproduce the experiments at the quantitative level. To understand the toroidal density distribution, toroidal cross-sections of the density profile at the outer mid-plane (z = 0.3 m) are shown in figure10 for the divertor (GIM11), top (GIM7) and mid-plane (GIM4) gas puff cases. In this figurethe toroidal positions of the ICRF antennas (in green color) and GIMs (white dashed lines) are also indicated. The orange line represents the position of the fast wave cut-off density during Figure 7. Poloidal cross-sections of the neutral density (nD2 for molecular and nD for atom) and electron density (ne) during divertor, top and mid-plane gas puffing, respectively. The poloidal cross-sections are chosen at toroidal positions where the GIMs are located. Nucl. Fusion 57 (2017) 056042 W. Zhang etal 15 confirm that combined gas puff can be effective in increasing the SOL density and the ICRF coupling globally. The validation of 3D simulations with the EMC3-EIRENE code and the experimental results in AUG [8] and JET-ILW make us confident that similar simulations can be carried out for ITER and other tokamaks. From our studies, it can be inferred that using middle plane gas valves closer to the powered ICRF antenna is the optimized way in terms of maximizing ICRF power coupling. We plan to do similar simulations for ITER and DEMO in the near future. Acknowledgments This work has been carried out within the framework of the EUROfusion Consortium and has received funding from the Euratom research and training programme 2014–2018 under grant agreement No 633053. The views and opinions expressed herein do not necessarily reflect those of the European Commission. References [1] GrahamM. etal 2012 Plasma Phys. Control. Fusion 54074011 [2] LamalleP. etal 2015 Radiofreq. Power Plasmas 1689030007 [3] StixT.H. 1992 Waves Plasmas (American Institute of Physics) [4] BilatoR. etal 2005 Nucl. Fusion 45L5–7 [5] JacquetP. etal 2016 Nucl. Fusion 56046001 [6] ZhangW. etal 2015 AIP Conf. Proc. 1689050006 [7] BobkovV. etal 2015 AIP Conf. Proc. 1689030004 [8] ZhangW. etal 2016 Nucl. Fusion 56036007 [9] JacquetP. etal 2012 Nucl. Fusion 52042002 [10] LercheE. etal 2015 J. Nucl. Mater. 463634–9 [11] SaibeneG. etal 1999 Nucl. Fusion 391133–56 [12] JoffrinE. etal 2014 Nucl. Fusion 54013011 [13] PinskerR.I. etal 2011 AIP Conf. Proc. 1406313 [14] PinskerR.I. 2011 37th EPS Conf. on Plasma Physics (Dublin, Ireland, 21–25 June, 2010) vol O4.124 (http://ocs. ciemat.es/EPS2010PAP/pdf/O4.124.pdf) [15] EkedahlA. etal 2015 Radiofreq. Power Plasmas 1689030013 [16] MaruyamaS. etal 2010 Proc. 23rd Int. Conf. on Fusion Energy (Daejeon) (Vienna: IAEA) CD-ROM file (http:// www-naweb.iaea.org/napc/physics/FEC/FEC2010/html/ fec10.htm) [17] MaruyamaS. etal 2012 Proc. 24th Int. Conf. on Fusion Energy (San Diego) ITR/P5-24 (http://www-naweb.iaea. org/napc/physics/FEC/FEC2012/papers/332_ITRP524.pdf) [18] FengY. etal 2004 Contrib. Plasma Phys. 4457–69 [19] ColasL. etal 2015 J. Nucl. Mater. 463735–8 [20] ZhangW. etal 2016 Plasma Phys. Control. Fusion 58095005 [21] Van EesterD. etal 2013 Plasma Phys. Control. Fusion 55025002 [22] LyssoivanA. etal 2012 Plasma Phys. Control. Fusion 54074014 [23] TripskýM. etal 2014 41st EPS Conf. on Plasma Physics (Berlin, Germany) P1.133 (http://ocs.ciemat.es/ EPS2014PAP/pdf/P1.133.pdf) [24] BobkovV. etal 2014 AIP Conf. Proc. 1580271 [25] JacquetP. etal 2011 Nucl. Fusion 51103018 [26] JacquetP. etal 2013 J. Nucl. Mater. 438S379–83 [27] ReiterD. etal 2005 Fusion Sci. Technol. 47172–86 [28] FengY. etal 2006 Nucl. Fusion 46807–19 [29] SharmaD. etal 2005 Nucl. Fusion 45825–36 [30] LuntT. etal 2012 Nucl. Fusion 52054013 [31] LuntT. etal 2015 J. Nucl. Mater. 463744–7 [32] LaoL.L. etal 1985 Nucl. Fusion 251421–36 [33] StangebyP.C. 2000 The Plasma Boundary of Magnetic Fusion Devices (Bristol: Institute of Physics Publishing) [34] LercheE. etal 2016 Nucl. Fusion 56036022 [35] JacquetP. etal 2014 Phys. Plasmas 21061510 Nucl. Fusion 57 (2017) 056042