Experiments and modelling of negative triangularity ASDEX Upgrade plasmas in view of DTT scenarios

The paper presents experimental and modelling results of a comparison of negative (NT) and positive (PT) triangularity ASDEX Upgrade (AUG) discharges using the plasma shapes presently foreseen in the DTT tokamak, under construction in Italy.


Introduction
The formation of an edge transport barrier (ETB, a region of reduced turbulent transport) in tokamak plasmas operating in H-mode [1] typically enhances the confinement performance.However, the H-mode is often accompanied by magnetohydrodynamic (MHD) instabilities called edge localised modes (ELMs).The ELM crash leads to energy and particle losses towards the divertor, compromising its lifetime.In recent years, many efforts have been done to explore alternative ELM-free scenarios, amongst which plasmas in negative triangularity (NT) shape (δ < 0), i.e. with the D-shape convexity directed towards the high field side of the toroidal magnetic field.Operations in NT have been shown to typically remain in L-mode at powers higher than the L-H threshold for positive triangularity (PT) scenarios, while also exhibiting relatively low anomalous transport levels within the separatrix, recovering the core performance of an H-mode [2][3][4].Indeed, previous experiments and theoretical studies on ECH heated TCV plasmas have pointed out the stabilizing effect of δ < 0 in trapped electron mode (TEM) dominated discharges [5,6], in which the NT electron heat transport is found to be halved obtaining the same PT kinetic profiles.In addition, improved confinement was recently demonstrated in experiments and non-linear gyrokinetic (GK) simulations for NT TCV plasmas with NBI heating, showing a turbulence suppression also for a mixed TEM and ion temperature gradient (ITG) regime [3,7,8].These results were confirmed by experiments and GK modelling on the DIII-D tokamak [9].
In recent years, NT configurations have been experimentally studied at the ASDEX Upgrade tokamak (AUG) [10,11], also to investigate the NT operation as an alternative scenario for the EUROfusion pilot project DEMO.In this context, this paper presents experimental results and numerical modelling of NT AUG plasmas in view of the design of DTT [12] (R = 2.19 m, a = 0.70 m, R/a = 3.1, B ⩽ 5.85 T, I p ⩽ 5.5 MA), a new D-shaped superconducting tokamak under construction in Italy, whose goal is to explore heat and particle exhaust strategies in a DEMO-relevant scenario.Since the NT equilibria that can be obtained in the AUG vacuum vessel are similar to those envisaged for DTT, in the last experimental campaign at AUG some pulses were performed to investigate NT plasmas with cross-sections achievable in the Italian tokamak.In fact, in both devices the lower triangularity is limited by geometrical and control coil constraints, and the minimum possible average δ is then restricted to values never below ~−0.2.A number of highly shaped PT discharges (as in DTT PT scenarios) were considered from the AUG database, with similar heating and safety factor q as the NT ones (figure 1).Unlike in TCV [8], NT plasmas in AUG tend to access the H-mode when operated in the common favourable ion ∇B drift configuration and moderate triangularities δ like those explored in this study [11].Therefore, a pulse has also been considered operating in unfavourable configuration to include at least one L-mode case with 1.5 MW of Electron Cyclotron Resonance Heating (ECRH): at higher powers the reversed field is still not sufficient to prevent the L-H transition in AUG.Results of the comparison of these NT and PT AUG pulses will be presented in this paper, both with experimental analysis and quasi-linear modelling using the 1D transport solver ASTRA [13,14] and the turbulent transport model TGLF-SAT2 [15].

Experimental setup and results
The experiments are performed in the AUG tokamak (R = 1.65 m, a = 0.50 m, R/a = 3.3, B ⩽ 2.5 T, I p ⩽ 0.8 MA), varying the average plasma triangularity −0.20 ⩽ δ avg ⩽ 0.46, plasma current I p = −0.6 and 0.8 MA, and elongation 1.61 ⩽ κ ⩽ 1.74, while the toroidal magnetic field strength B T = ±2.5 T is kept constant.Table 1 shows the main characteristics of NT and PT AUG plasmas with only ECRH power.The negative δ avg values are limited by the contour of the first wall, where δ avg = (δ up + δ low )/2.Indeed, while remarkable negative values of upper triangularity at the separatrix can be achieved (up to δ up ≈ −0.45), the lower triangularity (δ low ) is restricted by the wall structure above the divertor, leading to up-down asymmetric NT shapes and limiting the volume, see figures 1(a) and 2(g)-(j).In addition, the passive stabilizing loop (PSL) at AUG, which slows down the vertical displacement of the plasma, is less efficient in NT configurations [16] due to the larger distance from the plasma.More pronounced negative shapes tend to be less vertical stable.
The ECRH power is varied over time obtaining different heating levels during a single pulse.The gas puff in the vacuum vessel controls the line integrated plasma density, ranging from 2 to 7 • 10 19 m −3 .Low-Z species (B, C, N, O), together with W coming from the metallic wall and Fe/Ni from uncovered pipes, contribute to an effective charge of Z eff = 1.4 ± 0.3.In some cases, He is injected by piezo valves to infer the edge electron temperature T e and density  n e through the thermal helium beam diagnostic.However, this only affects the Z eff for a maximum of ∆Z eff ≈ 0.15.The EC power deposition, calculated using the paraxial beam tracing code TORBEAM [17], is radially peaked and localised inside ρ tor = 0.3 in every shot, where ρ tor := √ Φ/Φ sep and Φ being the toroidal magnetic flux.The experimental equilibria are obtained with the free-boundary equilibrium package IDE (Integrated Data analysis Equilibrium) [18], while the profiles through an integrated data analysis (IDA) [19] of different diagnostics, among which electron cyclotron emission (ECE) [20], Thomson scattering [21], and thermal lithium beam [22].Local and global performance of PT and NT plasmas are analysed looking at kinetic profiles, 0D parameters and volume integrated quantities in the central plasma region.Figure 3 illustrates the experimental electron kinetic profiles collected for the pulses with only ECRH power presented in this paper.CXRS measurement of ion temperature are missing since no beams are applied except in AUG #40473.First, the discharges with lower plasma current and input power (AUG #40647 PT and #40866 NT, I p ± 0.6 MA, P ECRH = 1.45 − 1.5 MW) are compared.Both pulses have constant line integrated density and MHD stored energy in the considered time interval, as displayed in figures 2(c) and (e).The PT and NT shots have an average triangularity δ avg of 0.39 and −0.17 and an elongation κ of 1.67 and 1.65, respectively.#40866 operates in unfavourable configuration, with the ion ∇B drift pointing away from the X-point and the toroidal magnetic field B T and plasma current I p reversed.The NT shot remains in L-mode during the whole phase (blue dashed lines in figure 2, left column), since the plasma requires more power for the L-H transition [10].This behaviour is highlighted by the absence of ELMs before ~2.8 s (Neutral Beam Injection (NBI) power on), as evidenced by the Mirnov coils signal in figure 2(k).The PT pulse operates in an ELM-free scenario known as EDA H-mode [23], an attractive plasma regime discovered for the first time in the Alcator C-mode tokamak [24].Although 1.45 MW of electron heating are sufficient for the pedestal formation (red profiles in figures 3(a)-(c)), no ELMs are detected.In figures 3(a)-(c) and 4(a) the NT (δ avg < 0) electron temperature T e and thermal pressure P e are shown to recover the PT central performance due to steeper temperature and density gradients in the region ρ tor ≈ 0.7 − 0.9.This behaviour persists at higher values of the EC auxiliary heating power, see figure 2(b).Indeed, the same trend is found comparing #36157 (δ avg = 0.40, κ = 1.70) and #40473 (δ avg = −0.14, κ = 1.61) with same current (I p = 0.8 MA) and ECRH power P ECRH = 2.9 MW.The line integrated density and MHD stored energy are constant also in this case (figures 2(d) and (f)).The D 2 gas puff is 25% more in the PT case.The NT discharge operates in favourable configuration and accesses the H-mode at low input power, P ECRH = 1.6 MW, as shown in figure 2(l).However, running with δ < 0 has shown to prevent the plasma    to develop high pressure pedestals, confirming the predictions reported in [25].The PT counterpart remains in an ELM-free H-mode up to 4.3 s and enters a Type I ELMy H-mode when 2.6 MW of EC auxiliary power are applied to the plasma.Due to lower pedestal pressure levels, the relative energy loss of the NT ELMs is smaller by at least a factor of 4 than the typical losses of Type-I ELMs appearing in the PT counterpart [11,26,27], as confirmed by electron temperature fluctuations at the edge (ρ pol = 0.96, ρ pol := √ Ψ/Ψ sep with Ψ the poloidal flux) and time-varying magnetic fluxes in the low field side of the machine, see figure 5.
Figures 3(d)-(f) and 4(b) show electron temperature, density, and pressure in the case of P ECRH = 2.9 MW.Although the difference in n e is more evident than in the previous case, the NT electron heat transport in the region 0.7 < ρ tor < 0.9 is such as to drive temperature gradients high enough to overcome the T e pedestal reduction and match the central electron thermal energy of the PT pulse.To investigate the role of the triangularity on the heat transport in the plasma core, it is helpful to look at the temperature logarithmic gradients R 0 /L Te = −R 0 d(logT e ), displayed in figure 6.With the same EC power, the PT H-mode features lower R/L Te inside the pedestal than the NT counterpart both in L-mode or in Hmode.This behaviour suggests again the presence of a region (ρ tor = 0.7 − 0.9) of reduced transport in these NT plasmas.Outside ρ tor = 0.95 the PT H-modes develop steeper gradients and higher R/L Te .
A comparison of global confinement properties is shown in figure 7. The NT plasmas operate in L-mode or exhibit lower pedestals than the PT H-modes: global parameters are expected to be lower, given the relevance of the edge volume in 0D global quantities.Indeed, the energy confinement time τ e and normalised beta β N are higher for the positive shapes.Nevertheless, if one refers to 'performances' as central performance, the volume averages of the electron pressure P e inside mid-radius become more relevant than  those 0D parameters.In this context, the NT H-mode and Lmode feature comparable/higher < P e > 0.5 values than the PT references.
Concluding, the NT shots recover the electron performance within ρ tor = 0.5 of the PT counterparts both remaining in Lmode or accessing the H-mode, featuring weaker ELM crashes in the latter case due to lower pedestal pressures.

Numerical modelling and transport predictions
Integrated modelling has been performed using the transport solver ASTRA (Automatic System for TRansport Analysis).The aim is predicting and investigating heat and particle transport in both NT and PT configurations, reproducing the experimental kinetic profiles of the ECRH heated plasmas analysed in section 2.1.Despite the high number of different species present in these discharges, two lumped impurities are calculated merging separately light and heavy elements.The neoclassical (NC) transport of main particles and light impurity is predicted by the fluid code NCLASS [28], while the new analytical model for collisional impurity transport FACIT [29] is used for the lumped heavy impurity.The latter better predicts the NC fluxes of high-Z atoms in a wider parameter space.The turbulent transport is calculated by trapped-gyro-Landaufluid (TGLF), a quasi-linear (QL) electromagnetic gyro-fluid model with shaped flux surfaces, using the SAT2 saturation rule.Predictions of the QL model TGLF in NT plasmas were already compared with the more precise gyrokinetic simulations in previous studies [30], finding a good agreement between the two approaches, making TGLF suitable for the study of NT geometry.
The main species are modelled up to the top of the pedestal in H-modes and up to ρ tor = 0.95 in the L-mode because TGLF does not include the pedestal physics and does not properly model the very edge region.The boundary conditions are set to match the experimental data.The impurities are predicted self-consistently up to the separatrix radius in all cases.The computation of ECRH deposition and initial profiles is based on heat flux calculations from TORBEAM and TRANSP [31].The power density is then kept fixed since these plasmas are simulated during the flat-top phase.Thus, electron temperature and density do not vary so much as to require a self-consistent resolution of the electron cyclotron heating.
The shape and position of the last closed flux surface have been prescribed as those calculated via IDE, while the axisymmetric equilibrium solver SPIDER [14,32] reconstructs the inner flux surfaces consistently.Sawteeth are not included in the modelling, and therefore the electron temperature profiles displayed here are referred to the levels just before the crash.TGLF predictions are not reliable inside ρ tor ≈ 0.2, especially in the case of a strongly peaked and central ECRH power density.Indeed, in such cases an experimental non-linear heat transport is expected, which makes the resulting T e profiles flatter than those predicted by the modelling.Therefore, an additional electron diffusivity χ e has been added inside this region to match the experimental gradients.The investigation of this behaviour is outside the scope of this work.
Figures 8 and 9 show the comparison between experiments and modelling results in the case of ~1.5 MW and 2.9 MW of ECRH.Looking at the predicted T e , n e , and P e , the simulations catch most of the core physics seen in the experiments both for NT and PT, showing a reasonable agreement between dashed and solid lines.The only available measurement of T i is also reproduced.
The ion central performance has been evaluated using the T i and P i (figures 8(a), (c) and 9(b), (d)) profiles from ASTRA/TGLF.The expected fusion performance, which is proportional to P 2 i ( [33], equations (1.4.1) and (1.5.4)), has been compared by calculating the volumetric averages of P 2 i and T i inside mid-radius (see figure 10), where the major fraction of the fusion power is expected.The good properties observed in NT are mainly attributable to reduced heat transport in the electron channel.Indeed, while the electron temperature gradients are steep enough to balance the difference in n e and lead to better electron central pressures, the modelling predicts similar T i gradients and central levels.In this context, the density plays a crucial role in terms of electron-to-ion power and ion pressure, and thus the fusion performance of the PT high density overcomes the one calculated for the NT reference at 2.9 MW.The performance of  NT seems to degrade as the ECRH power increases.However, the observed degradation is rather due to the difference in density than the increment in ECRH power.In fact, #40036 PT and #40869 NT (see section 3.1) feature similar density and comparable ion performance with 7 MW of total power injected.
If one looks at the predicted R/L Te in figure 11, the model seems to partially overestimate the NT electron heat transport in the region of interest (ρ tor = 0.7 − 0.9), even if the NT logarithmic gradients are correctly higher than those predicted for the PT counterparts.TGLF uses a Miller analytical equilibrium [34], which approximates the magnetic flux surfaces as up-down symmetric.To do that, in the case of an asymmetric LCFS like in these NT AUG shots, the outer surfaces (where the triangularity is more pronounced) lose part of the tilt with respect to the vertical axis (figure 12).This results in a reduction of the upper negative triangularity effects in the outer plasma.
TGLF predicts ITG modes as the dominant unstable drift for most of the radial points, as predicted for the DTT full power scenario in both δ configurations [30,35].A numerical  test by changing only the shape is done to discern the impact of the triangularity.The numerical experiment consisted in flipping the LCFS of AUG #40866 and #40647 symmetrically with respect to the magnetic center of the separatrix (figure 13) and running the simulations with the other parameters and boundary conditions unchanged.If one looks at figures 14(a),(b) and (d), the NT cases (i.e. the original #40866 and #40647 with flipped boundary) have higher R/L Te and steeper T e,i gradients than the PT counterparts in the region where the triangularity is expected to count.The action of flipping the separatrix shape has therefore a non-marginal beneficial effect on the heat transport.However, the difference between NT and PT transport levels cannot be entirely explained by the geometry impact between ρ tor = 0.7 and the outermost simulated point.Indeed, the artificial #40866 with positive δ still overcomes the central performance of the original #40647 and vice versa.Thus, the gain of running in NT seems to be also attributable to other features, among which geometry effects outside the simulation domain (i.e. in the pedestal and the Scrape-Off-Layer), that impact the TGLF predictions in the role of boundary conditions.Furthermore, the larger beneficial effect is observed when flipping the PT shape because the artificial NT has an up-down symmetric LCFS with a strong negative delta of ≈−0.4,which is not even limited by the Miller approximation.

Experimental setup and results
The NT option currently under investigation for the DTT full power scenario (B t = 5.85 T, I p = 4 MA) considers both electron and ion heating.Thus, a comparison between positive and negative triangularity AUG plasmas with mixed NBI+ECRH heating has been included in this work.The main features of the experiments presented in section 2 remain valid for these cases.
Despite #40869 operates in the unfavourable configuration of the ion ∇B drift, all four discharges access the Hmode.Their LCFS are displayed in figure 15.While the two NT pulses and #40036 have shapes like those investigated in section 2 (i.e.NT with a low average delta and PT with a high one), #40036 features a moderate PT value of 0.27 and a shape different from the one foreseen for DTT.However, its low density makes the latter a case of interest, considering that most of the other high-δ PT plasmas in this work almost double the density of their NT counterparts.In this specific case, the high density of #39988 is justified by higher D 2 puffing and better performance due to its high δ.
Figures 16 and 17 compare experimental T e , n e , P e , T i , and P i .The ion data are obtained using the charge exchange recombination spectroscopy (CXRS) [36,37].As reported for pure ECRH shots in section 2, the NT plasmas are shown  to overcome the central electron temperatures and pressures of the low-δ PT reference (δ avg = 0.27) due to steeper gradients outside ρ tor = 0.7.Again, the ion temperature gradients look the same.However, #40869 and #40470 in NT have performance good enough to recover the total thermal pressure of #40036.P i and P tot are calculated assuming a mix of impurities consistent with the experimental Z eff and applying the quasi-neutrality since only boron and tungsten concentrations are well known in these shots.The experimental error bars are therefore not available for the displayed profiles.
Global 0D parameters and volume averages inside midradius have been compared (figure 18).As expected, β N and τ E of the low-density PT plasma are comparable with those of the counterparts in NT.Moreover, #40869 (black) performs better than the other shot with δ < 0 (blue) and overcomes the central and even fusion performance (P e and P 2 i integrated within ρ tor = 0.5) of the low-δ PT reference (red).
On the other hand, comparing the two NT shots with the high-density PT plasma necessitates a different discussion, similar to what is reported for the ECH-only shots at 2.9 MW.With similar ion temperature and consistently lower density, #40869 in reverse field configuration features worst fusion performance (lower < P 2 i > 0.5 ) than the PT reference, even if it still recovers the electron and total pressures inside mid-radius.
With the same power, the NT H-modes feature higher R/L Te inside the ETB (0.75 < ρ tor < 0.85) if compared with the PT cases.A region of reduced electron transport in AUG NT plasmas seems to exist also in the case of mixed ion/electron heating, as shown in figure 19.
Despite all the analysed shots exhibit ELMs, the fast time resolution ECE measurement of T e at the edge confirms the trend found in section 2, showing a gain in operating in regimes with a negative delta.Indeed, the temperature fluctuations associated with those crashes are consistently lower in NT (figures 20 and 21), as well as the time-varying magnetic field detected in the lowfield side using the Mirnov coils signal displayed in figure 22.

Numerical modelling and transport predictions
The quasi-linear simulations are performed using the same suite of codes and models exploited to predict the ECRH-only cases.Again, the plasma boundaries are reconstructed using the package IDE and the inner flux surfaces predicted with SPIDER.The NBI fast ion deposition is previously calculated with the module NUBEAM [38] (included in TRANSP) and kept fixed as done for the EC power density.The contribution of the rotation to transport has been included in the modelling via the E × B in TGLF and the effects on impurities in FACIT.In fact, poloidal asymmetries can arise in the heavy impurity distribution function with a non-null toroidal velocity, leading to substantial changes in the neoclassical pinch [29].In the case of high-Z species, for which the neoclassical transport can compete with the turbulent one, the effects of inhomogeneous densities cannot be neglected.Therefore, the use of FACIT has been mandatory since NCLASS considers homogeneous distribution functions.
The results are displayed in figure 23.In general, ASTRA/TGLF predicts the experiments qualitatively with reasonable accuracy.The ion temperatures are reproduced within uncertainties in the region of interest (ρ tor = 0.6 − 0.9), while they diverge from the experimental profiles in the inner plasma core.Again, the use of Miller equilibrium in TGLF leads to a partial overestimation of the NT electron heat transport between ρ tor = 0.7 − 0.9.Moreover, the PT T e prediction is higher than the experimental profile, especially around mid-radius.In contrast to what we have observed for other PT discharges, this low-δ case has up-down asymmetric flux-surfaces (δ low = 0.45 and δ up = 0.09), see figure 24.This characteristic seems to limit the prediction capabilities of ASTRA/TGLF even for a PT scenario.
As expected, TGLF predicts ITG modes as the dominant unstable drift for both delta configurations, confirming the main results obtained through linear gyrokinetic analyses in [10].

Conclusions
Experimental and numerical results of NT plasmas in the AUG tokamak are presented.In particular, pulses with NT shapes like those foreseen for the DTT device have been investigated, comparing their local and global performance with PT scenarios featuring similar heating, safety factor, and elongation.In AUG, the contour of the first wall and the forces due to the coil currents limit the triangularity values in the case of δ < 0, and plasmas in favourable B-field configuration tend to access the H-mode at low input power though operating in NT.Therefore, a pulse in unfavourable configuration has also been considered to include at least one L-mode case.
Discharges in pure ECRH and mixed NBI+ECRH have been analysed.The experiments have shown good confinement properties of these NT AUG plasmas in both heating regimes.The NT electron pressures are shown to recover the PT counterparts due to reduced heat transport in the electron channel and higher electron temperature and density gradients in the region ρ tor = 0.7 − 0.9, leading to similar or higher volumetric averages of T e and P e inside mid-radius.On the other hand, the ion temperatures feature similar gradients and central levels.Thus, the ion pressures and related fusion performance look higher for those high-density PT plasmas with strong triangularity and high gas puff, while they are comparable in the case of PT and NT shots with similar densities.Moreover, plasmas with δ < 0 exhibit weaker individual ELM energy losses due to lower pressure pedestals, regardless of the amount and type of power injected.
Integrated modelling has been performed using the transport solver ASTRA and the quasi-linear turbulent model TGLF-SAT2.The simulations reproduce the general aspects of the experiments, predicting the kinetic profiles with reasonable accuracy.However, the use of Miller equilibrium partially limits the predictive capabilities of TGLF for scenarios with an up-down asymmetric shape like some of those analysed in this work.A numerical experiment to isolate the impact of the geometry by symmetrically flipping the boundary has shown a beneficial effect of the negative delta on heat transport.Nonetheless, it justifies only part of the role of NT observed in AUG experiments.Again, the Miller parametrisation could be responsible for the transport underestimation when an NT shape is flipped into a PT one.Additionally, a second contribution could come from NT beneficial effects in the SOL [8,39], which impact the boundary conditions that remain unchanged in this numerical exercise.
The confinement results presented in this paper demonstrate some practical gains of running plasmas in NT, despite the limited triangularity values.The outcomes are therefore fairly encouraging in view of the DTT device, which is meant to operate in the same range of NT.

Figure 1 .
Figure 1.(a) An example of poloidal cross-sections obtained for PT and NT AUG plasmas compared with the shapes presently foreseen for DTT.(b) Safety factor q from IDE equilibrium reconstruction (see section 2) for some AUG PT/NT scenarios.

Figure 4 .
Figure 4. Electron thermal pressure in the region ρtor = 0.6 − 1, an enlarged view of the outer profiles of figures 3(c) and (f).

Figure 5 .
Figure 5. Left (a): time-varying magnetic fluxes detected by the Mirnov coils in the low field side.Right (b): electron temperature fluctuations at ρtor = 0.96 measured by the ECE diagnostic.

Figure 6 .
Figure 6.Electron temperature logarithmic gradients compared for NT L-mode, PT H-mode, and NT H-mode in pure ECRH shots.

Figure 7 .
Figure 7. Global parameters and volume integrated quantities calculated for ECRH heated plasmas.On the x-axis powers are expressed in MW.(a) Electron pressure integrated within ρtor = 0.5.(b) Upper, lower, and average triangularity of the LCFS.(c) Normalised beta.(d) Energy confinement time.

Figure 8 .
Figure 8. ASTRA/TGLF predictions compared with experiments for the ECH-only cases.Electron and ion temperatures T e,i (a), (c) and electron density ne (b), (d).

Figure 10 .
Figure 10.Predicted ratios between PT and NT fusion and ion performance using P 2 i and T i averaged within ρtor = 0.5.

Figure 13 .
Figure 13.(a) Comparison between the original LCFS of #40866 NT 1.5 MW and the same boundary flipped with respect to the center of the separatrix.(b) Same comparison for #40647 PT 1.45 MW.

Figure 14 .
Figure 14.The effect of flipping the geometry on electron temperature Te (a), electron temperature normalised gradient R/L Te (b), electron density ne (c), and ion temperature T i (d).

Figure 17 .
Figure 17.Experimental electron pressure from IDA and ion pressure of AUG plasmas with NBI+ECRH.The ion pressures are evaluated assuming a mix of impurities consistent with the experimental Z eff .

Figure 18 .
Figure 18.Global parameters and volume integrated quantities calculated for ECRH+NBI cases.(a) Electron thermal pressure integrated within ρtor = 0.5.(b) Average triangularity of the separatrix.(c) Ion pressure squared integrated inside mid-radius.(d) Energy confinement time.FW: forward toroidal magnetic field.RV: reverse toroidal magnetic field.

Figure 20 .
Figure 20.Electron temperature fluctuations detected by ECE diagnostic.The visible ELM cycles in red is detected in the PT pulse.

Figure 21 .
Figure 21.Time-averaged fluctuations of the electron temperature at the edge measured by the radiometer (ECE diagnostic).

Figure 22 .
Figure 22.Time-varying magnetic field measured by the Mirnov coils in the low-filed side of the machine.Each peak corresponds to an ELM crash.

Figure 23 .
Figure 23.ASTRA/TGLF predictions compared with experiments for ECRH+NBI AUG plasmas.From the left: electron temperature Te (a), electron density ne (b), and ion temperature T i (c).

Table 1 .
Main characteristics of NT and PT AUG plasmas investigated in this work in pure ECRH power.

Table 2 .
Main characteristics of NT and PT AUG plasmas with mixed ECRH+NBI power.