First principle integrated modeling of multi-channel transport including Tungsten in JET

Figures 4–9 present time traces of several plasma parameters which are self-consistently and simultaneously evolved in the simulation: density, temperature, rotation, current profiles as well as impurity content and radiation. For each parameter, the JETTO-SANCO-QuaLiKiz-NEO-PENCIL-FRANTIC-ETB simulation is shown in magenta, and compared with experimental measurements when available. Timetraces are shown at three ρ positions: 0.1, 0.4, and 0.75.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 4 shows the time evolution of electron density. The density prediction must be accurate in order to simulate properly the W transport, due to the main ion density gradient dependence of the W neoclassical transport (see equation (2) of [35]). On figure 4(a), close to the center, the predicted density increases smoothly. The HRTS measurements, while also globally increasing, are impacted by the sawteeth. The simulation does not take sawteeth into account, therefore the local gradients are not well captured at all times. For the two other radial positions, the predicted density does not vary much, and mostly stays within experimental uncertainties.

Figure 5 shows the electron temperature. Similarly to the electron density case, the simulation successfully captures the decreasing trend of central ECE measurements on figure 5(a), but expectedly misses the sawtooth crashes. Around mid-radius (figure 5(b)) and close to the pedestal (figure 5(c)), the experimental temperature remains quite steady and the predictions lie within experimental uncertainties over time.

Ion temperature and toroidal rotation are not compared with experimental data in the center (figures 6(a) and 7(a)) due to lack of charge exchange data there. At $\rho=0.4$ on figure 6(b), the predictions tend to underestimate the charge exchange measurements by at most 15% (calculated from the lower bound of the error bar). A potential source for this T_i underestimation is the lack of electromagnetic stabilization of ITG turbulence [36] in the electrostatic QuaLiKiz model. Indeed, the pulse studied here is a high-performance hybrid H-mode, where electromagnetic stabilization is typically more prominent due to high-β and reinforced by a high fast ion fraction, as shown by nonlinear gyrokinetic simulations for a JET hybrid pulse in [37].

Toroidal rotation predictions mostly lie within experimental uncertainties, except at early times at $\rho=0.75$. Overall, the experimental toroidal velocity remains quite constant over time. Note that the plasma rotates up to 300 km s⁻¹, which leads to a W Mach number up to 3 leading to strong poloidal asymmetries.

The predictions of safety factor evolution are shown on figure 8. The self-consistent current diffusion equation solved in JETTO leads to these predictions of q. This predicted q differs slightly from the EFIT reconstruction constrained by kinetic pressure and Faraday angles [32].

Figure 9(a) shows the average Z_eff time evolution and figure 9(b) shows the total radiated power over time. The Z_eff globally increases with time, which can be explained by the fact that W contribution to Z_eff increases also with time. The line integrated radiated power does not increase over time, nor in the experiment neither in the simulation. The simulation does not account for ELMs, therefore it does not capture the regular bursts reported on the bolometer signal. The fact that the W core accumulation is not visible on the line integrated bolometer signal is due to the fact that the W cooling factor is maximum around T  =  1.5 keV [33], hence off-axis for the pulse studied here.

Figure 10 shows the time evolution of the W content. The simulation values are renormalized to $\langle n_{\rm W, sim}(0.4)/n_{\rm W, exp}(0.4)\rangle$, i.e the time average of the ratio of the simulation value over the measurement value at $\rho=0.4$ and $\theta=0$. Importantly, note that flux surface averaged and 2D W profiles are compared without renormalization on figures 15–17. The initial W content was estimated from the total radiation shown on figure 9(b). Due to uncertainties, some experimental values are unavailable. In the very core, $\rho=0.05$, the simulated W density slowly and regularly increases, with a factor 3 increase between 6 s and 7 s. The W density inferred from SXR-UV measurements [20] accumulates a little at 6.3 s before being flushed out by the second sawtooth. Between the second and the third sawtooth, from 6.7 s, the core W rises by a factor 10. The JETTO-SANCO-QuaLiKiz-NEO-PENCIL-FRANTIC-ETB simulation, which does not account for sawteeth, captures the core W accumulation trend. The predicted W content at $\rho=0.4$ and $\rho=0.8$ are compared to the measured ones in figures 10(b) and (c). Note that in a JETTO-NEO simulation, where all the channels where run interpretatively but W transport which was predicted by NEO, a much better agreement of the W time evolution was obtained. This demonstrates that the mismatch in the temporal evolution of W is due to the difference in the predicted versus interpreted density, rotation and temperature profiles, not due to the W transport modelling.

Zoom In Zoom Out Reset image size

To deepen the analysis of the quality of the JETTO-SANCO-QuaLiKiz-NEO-PENCIL-FRANTIC-ETB predictions, the next subsection focuses on several profiles at different time slices, and 2D poloidal cuts of the W distribution.

4.2.1. Electron density.

In order to further validate QuaLiKiz predictions, figure 11 shows electron density profiles at three different times: 6.2 s after 0.7 s of simulation (figure 11(a)), 6.8 s after 1.3 s of simulation (figure 11(b)) and 7 s, after 1.5 s of simulation (figure 11(c)). Both HRTS and LIDAR are shown on figure 11 but we shall compare the simulation with HRTS only because of its higher time and space resolutions.

Zoom In Zoom Out Reset image size

On figure 11(a) at 6.2 s, QuaLiKiz predictions in magenta lie within experimental uncertainties of the HRTS in blue for the whole radius, except close to the axis where QuaLiKiz misses the local variations. The predicted pedestal is modeled by the ETB which has been tuned to match the experimental data. On figure 11(b) at 6.8 s, QuaLiKiz prediction lies within experimental uncertainties from R  =  3.3 m outward. But from R  =  3.3 m inward, the experimental density dropped because of a sawtooth, therefore QuaLiKiz overestimates the central electron density. The experimental pedestal is stable and well captured by the edge transport barrier modeling. On figure 11(c) at 7 s after 1.5 s of simulation, HRTS shows strong local gradients, especially close to the axis at R  =  3.1 m. The experimental pedestal is slightly shifted outward, therefore the Edge Transport Barrier no longer lies within experimental uncertainties. QuaLiKiz captures the global increase of the electron density and lies within experimental uncertainties. However it does not fully reproduces the stiff central gradient shown by the measurements. This impacts the W transport as seen in the next section, and could be improved in the future by including a sawtooth model.

4.2.2. Electron and ion temperatures.

The electron and ion temperature predictions are then compared with experimental data. The electron temperature profiles are first shown on figure 12: at 6.2 s after 0.7 s of simulation (figure 12(a)), at 6.8 s after 1.3 s of simulation (figure 12(b)) and at 7 s after 1.5 s of simulation (figure 12(c)). QuaLiKiz predictions are in magenta solid line, and the HRTS measurements in blue.

Zoom In Zoom Out Reset image size

On figure 12(a) at 6.2 s, QuaLiKiz predictions of T_e in magenta lie within experimental uncertainties of the HRTS measurements in blue accross the whole radius, except at R  =  3.3 m where QuaLiKiz misses a bump in T_e and slightly underestimates the measurements. On figure 12(b) at 6.8 s, the HRTS shows a global decrease of the electron temperature, while keeping the central peaking and the bump at R  =  3.3 m. The experimental pedestal remains unchanged and is well reproduced by the Edge Transport Barrier. QuaLiKiz predicts the global decrease of the electron temperature, but still misses the bump at R  =  3.3 m. QuaLiKiz predicts the central peaking and slightly overestimates it. On figure 12(c) at 7 s after 1.5 s of simulation, the bump at R  =  3.3 m disappears while a drop appears at R  =  3.5 m. QuaLiKiz predictions barely change compared with 12(b). Therefore it misses the drop at R  =  3.5 m while staying within experimental uncertainties. QuaLiKiz still overestimates the very central temperature.

Now the ion temperature profiles are shown on figure 13: at 6.2 s after 0.7 s of simulation (figure 13(a)), at 6.8 s after 1.3 s of simulation (figure 13(b)) and at 7 s after 1.5 s of simulation (figure 13(c)). QuaLiKiz predictions are in magenta solid line, and the charge exchange measurements are in black for most of the radial points, and blue for the pedestal.

Zoom In Zoom Out Reset image size

On figure 13(a) at 6.2 s, QuaLiKiz predictions in magenta lie within experimental uncertainties of the Charge Exchange measurements for the whole radius. On figure 13(b) at 6.8 s, the Charge Exchange measurements show a global increase of the ion temperature, up to 50% at R  =  3.2 m. QuaLiKiz predictions remain almost unchanged and therefore underestimates the measurements by a maximum of 15% (calculated from the lower bound of the error bar). The possible reasons for this underestimation were developed in the previous section. On figure 13(c) at 7 s after 1.5 s of simulation, measured ion temperature slightly decreases at R  =  3.25–3.4 m. QuaLiKiz predictions barely change compared with figure 13(b). Therefore it still underestimates the measurements at R  =  3.25–3.4 m.

4.2.3. Rotation.

The last plasma parameter to study is the toroidal rotation, shown on figure 14: at 6.2 s after 0.7 s of simulation (figure 14(a)), at 6.8 s after 1.3 s of simulation (figure 14(b)) and at 7 s after 1.5 s of simulation (figure 14(c)). QuaLiKiz predictions are in magenta solid line, and the Charge Exchange measurements are in black for most of the radial points, and blue for the pedestal.

Zoom In Zoom Out Reset image size

On figure 14(a) at 6.2 s, QuaLiKiz predictions in magenta lie within experimental uncertainties of the Charge Exchange measurements for the whole radius. The central part also lacks experimental measurements. On figure 14(b) at 6.8 s, the charge exchange measurements show a slight global decrease of the toroidal rotation, especially at R  =  3.5–3.65 m. QuaLiKiz predicts the global decreasing but overestimates the strongest decrease of the measurements at R  =  3.5–3.65 m. Finally, on figure 14(c) at 7 s after 1.5 s of simulation, measured velocity profile moves slightly. QuaLiKiz predictions barely changed compared with 14(b). Therefore it slightly overestimates the measurements at R  =  3.3–3.7 m and the edge transport barrier model, which does not evolve, misses the pedestal.

To analyze more precisely the time evolution of the W profile, figure 15 shows the Flux Surface Averaged W profiles at t  =  6.2 s after 0.7 s of simulation and t  =  7.1 s after 1.6 s of simulation. The experimental data is not available from $\rho=0.7$ outwards where the SXR signal is too weak to constrain the W profile reconstruction. Note that here, unlike for figure 10, the profiles are not renormalized one to the other. And also note that, unlike previous works [6, 7], where each 2D n_W prediction was independently normalised to SXR, here the W content is, for the first time predicted in absolute value.

Zoom In Zoom Out Reset image size

At 6.2 s on figure 15(a), the simulation overestimates the W content inferred from measurements by a factor 2 at mid-radius and up to a factor 10 in the very center where, according to the experimental data, the W did not migrate yet. This discrepancy probably comes from the initial condition which constrains the W profile to be homothetic to the electron density profile. At 7.1 s, after 1.6 s, on figure 15(b), the simulated central W content is very similar to the data inferred from SXR measurements, hence capturing the W core accumulation. The simulation still slightly overestimates the measurements from $\rho=0.2$ outward.

Figures 16 and 17 show 2D maps of the W density at two different times: t  =  6.2 s after 0.7 s of simulation and t  =  7 s after 1.5 s of simulation. W densities estimated from SXR-UV measurements are on the right, predictions are on the left.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

On figure 16, one can see that the W density, both measured and simulated is in the $10^{15}~{\rm m}^{-3}$ range. The initial amount of W was homothetic to the electron density profile, which is very different from experiment (see figure 15(a)). Therefore, in the simulated profiles, there is W in the very core from the start, unlike the experiments. Concerning poloidal asymmetries, at 6.2 s, both the measured and the simulated profiles show strong poloidal asymmetries, although they are weaker in the predicted profiles.

At the end of the simulation, after 1.5 s at t  =  7 s, the estimation from measurements on figure 17(b) shows that most of the W moved towards the center and accumulated. On the simulation, most of the W kept moving towards the center, but not fast enough and a significant W amount is still present at mid radius (see figures 10(b) and (c)). However the time evolution of the W behavior definitely shows a trend of core accumulation.

Three phenomena can explain why the simulation does not fully succeed in transporting all the W to the plasma center. The first explanation is that the simulation does not model sawteeth and therefore some transport mechanisms can be missed. The second explanation is that the initial W profile, with some W already in the center and at mid-radius unlike the experiment, remains present even after a few confinement times. The third explanation could be that QuaLiKiz, as seen on figure 11 globally captures the global central density peaking but does not fully reproduce all the local gradients, especially at mid-radius and in the center. This could lead to underestimated neoclassical transport and therefore a weaker central accumulation.

Overall, the accuracy of the QuaLiKiz predictions of the main plasma profiles, especially electron density and rotation, allows NEO and QuaLiKiz to correctly estimate the W neoclassical transport and therefore reproduce the W central accumulation trend even though more work is required to capture all the experimental features of the background profiles, in particular the sawteeth impact on them.

According to equation (2) of [35], in the case of an inward W particle flux, a larger main ion density gradient leads to a larger inward W neoclassical flux. In the case of 82 722, the particle fuelling is stronger in the central part (see figure 18(a)), causing the particle density to peak (figure 4(a)).

Zoom In Zoom Out Reset image size

In order to confirm the link between central particle fuelling and central W accumulation, two new simulations are set: one with the particle injection artificially set to zero in PENCIL settings at all times; and the other with partial off-axis particle source at all times. In all cases, only the PENCIL NBI particle source is modified, the NBI heating and angular momentum input are kept identical to the original simulation. Figure 18(a) illustrates the NBI particle source profiles at 6.5 s for the three cases.

Figure 18(b) shows the W density Flux Surface Averaged over time at ρ  =  0.05. The shaded section on figure 18(b) from t  =  5.5 s to t  =  5.7 s corresponds to the first simulated confinement time, needed for the simulation to move away from initial conditions. The removal of the central density source completely mitigates the W accumulation phenomenon. With a 45 % reduction of the central particle source, the central W density is reduced by 42%. Figure 18(c) shows the W density Flux Surface Averaged profile at t  =  6.5 s. Without the central particle source, there is almost no W in the central part, most of the W is located in the outer half. The W density profiles of the reference simulation and the simulation with partial off-axis particle fuelling are quite similar from ρ  =  0.4–1, but in the center, the partial off-axis particle fuelling reduces the W central accumulation by 45%.

When switching off, or reducing the particle source, the background plasma profiles are impacted as well. This is illustrated on figure 19 for the density, temperatures and rotation profiles at t  =  6.5 s. As expected, the removal on the central particle fuelling annihilates the central electron density peaking. The reduction of central fuelling by $45\%$ in red leads to a reduction of the central density peaking by only $25\%$. Since without central particle fuelling, W does not accumulate, larger temperatures are obtained in absence of core particle fuelling.

Zoom In Zoom Out Reset image size

Overall, this study demonstrates the deleterious correlation between NBI central particle fuelling and the W central accumulation. This is encouraging for devices such as WEST or ITER, with off-axis particle fuelling.

The other parameter impacting significantly the W neoclassical transport is the rotation. The impact of poloidal asymmetries in the W neoclassical transport is accounted for in the geometrical terms P_A and P_B in equation (2) of [35], and their range of applicability is studied in [38]. In this specific rotating JET-ILW pulse, these geometrical factors enhance the neoclassical convection up to a factor 40. In order to study the impact of rotation on the W central accumulation, a simulation is set with the NBI angular momentum set to zero while keeping identical the heat and particle sources. The resulting torque, and core W density evolution are illustrated on figure 20.

Zoom In Zoom Out Reset image size

On figure 20(b), during the first confinement time in the shaded area, for the reference simulation in magenta, the central W content drops before increasing again. Indeed W density profile equilibration time scale is much shorter than the energy confinement time scale. The removal of the NBI angular momentum makes the W transport much more sensitive to the initial condition. However, for the simulation with no torque, the W content remains stable and then increases. This makes the two simulations not directly comparable. In order to remove the effect of the first confinement time on W density, figure 20(c) shows the timetrace of W density normalized to its value at the end of the first confinement time, at t  =  5.7 s. Without NBI angular momentum, W central content doubles over time, while in presence of toroidal rotation, the W density increases by a factor 10. This clearly illustrates that the toroidal rotation plays a role in the W central accumulation process.

The removal of the NBI angumar momentum causes the two simulations to have very different early phases and therefore completely different time evolutions. We shall focus on the first confinement time in order to understand the mechanisms leading to such a big difference.

As mentioned earlier, the initial W density profile is homothetic to the electron density profile, therefore some W is already present in the center at the start. In absence of NBI torque, the W is equally poloidally distributed along each flux surface which reduces significantly its neoclassical inward convection. Figure 21 shows the W transport coefficients at the end of the first confinement time, at t  =  5.7 s, time averaged over 0.1 s to smooth the QuaLiKiz predictions.

Zoom In Zoom Out Reset image size

As expected, the removal of the torque reduces by a factor P_A, the neoclassical diffusion and convection (on figures 21(c) and (d)). The neoclassical convection is strongly reduced both in the very core and in the pedestal region.

The absence of torque also reduces the turbulent effective convection and diffusion for $\rho>0.5$ by up to a factor 10. Note that the outputs to JETTO are either a $D_{\rm turb, \, ef\,\!f}$ or a $RV_{\rm turb, \, ef\,\!f}$, depending on the direction of the transport and the size of the gradient. Therefore there are radial regions of zero $D_{\rm turb, \, ef\,\!f}$ that have non-zero $RV_{\rm turb, \, ef\,\!f}$, and since it is averaged over 0.1 s, in some cases both $D_{\rm turb, \, ef\,\!f}$ and $RV_{\rm turb, \, ef\,\!f}$ can be at play. The impact of poloidal asymmetries on W turbulent transport is accounted for in QuaLiKiz based on [39], as described in [16]. The impact of centrifugal effects on the various components of the turbulent convection (thermodiffusion, rotodiffusion, pure convection) results of a complex compensation of the different components (see [39]) leading, in this case, to a reduced turbulent W transport in absence of torque.

Overall, in absence of NBI torque, both turbulent and neoclassical W transport are reduced. As a consequence, W is no longer flushed out from the central zone of the plasma over the first confinement time. During the rest of the simulation, it is weakly transported to the center by residual neoclassical convection, but the W amount transported is negligible in absence of NBI angular momentum compared with the case with NBI torque (see figure 20(c)).

When removing the NBI torque, the turbulent main ions heat and particle fluxes are also affected (see [40] for details on the $E\times B$ implementation in QuaLiKiz). Figure 22 shows the heat fluxes profiles from JETTO simulations (reference and without rotation) at t  =  5.7 s, after the first confinement time. When setting to zero the torque, the $E\times B$ shear is largely reduced [41]. Due to a weaker $E\times B$ shear, the heat coefficients are larger in absence of torque. As a consequence, the electron and ion temperatures are lower in, as seen on figures 23(b) and (c). Note that the electron density profile on figure 23(a) is weakly impacted, except for the central density peaking which is slightly higher in absence of torque. The central radiation on figure 23(d) is larger in absence of torque since the W is not flushed out.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In summary, the removal of the NBI angular momentum strongly reduces the W neoclassical convection, as expected, and also its turbulent transport (see figure 21). But removing the torque leads to more unstable background plasma heat and particle transport (see figures 22(a) and (b)). Initially, the removal of the rotation is deleterious for both the energy confinement (increased turbulence for main ion and electrons) and the central W content since W is no longer flushed out. But once the impact of the initial condition removed, it appears that NBI torque has a negative impact on the W central accumulation (see figure 20(c)). Therefore lower torque experiments, such as WEST or ITER, should be less prone to core W accumulation.

When removing the W from our reference simulation the radiation level is almost reduced to zero. Therefore to isolate the impact of radiation from the impact of the W itself, a simulation as our refernce case but with the radiation forced to zero is run.

It is interesting to note that the energy content of the Be only case and W-Be but no radiation case are very similar (within 1%), see figure 25(d). When comparing the effective ion heat diffusivities, removing the radiation from the reference case is destabilizing for $\rho>0.6$, see figure 25(a). A similar impact is seen also on the electron heat effective diffusivity, figure 25(b), as well as on the ion particle effective diffusivity, figure 25(c).

Zoom In Zoom Out Reset image size

In order to have a better understanding of the stabilization effect, the timetraces of ion and electron temperatures, the ratio $T_{\rm e}/T_{\rm i}$ and the ion heat diffusivity are shown on figure 26 for the position ρ  =  0.7.

Zoom In Zoom Out Reset image size

The curves split in two groups: the reference simulation with W and radiation on one side (magenta), and the simulations without radiation either with or without W on the other side. Without radiation, as expected, the electron temperature on figure 26(a) is higher. The ion temperature on figure 26(b) is also impacted and is lowered. The modification in the temperature impacts the ratio $T_{\rm e}/T_{\rm i}$ shown on figure 26(c), which is higher for the simulations without W radiation.

The enhanced electron temperature, combined with lower ion temperature leads to enhanced $T_{\rm e}/T_{\rm i}$, which is known to increase ITG dominated turbulence [27]. Therefore, the destabilization effect could be explained by this mechanism: the removal of W causes the radiation level to be significantly reduced, leading to an enhanced $T_{\rm e}/T_{\rm i}$, causing the turbulence to be increased. In order to validate this explanation, a QuaLiKiz stand-alone simulation is run, scanning the electron temperature. All the other inputs are taken from the JETTO reference simulation parameters at t  =  6.5 s and ρ  =  0.7. The profile values entered as input in QuaLiKiz are taken on each JETTO flux surface (except the W density taken at the outboard mid-plane) and the gradients are calculated along $(R_{\rm max}-R_{\rm min})/2$, where R_max (respectively R_min) is the maximum (respectively minimum) major radius of each flux surface. The ion and electron heat effective diffusivities and the ion particle effective diffusion are shown on figure 27. The red diamond corresponds to the $T_{\rm e}/T_{\rm i}$ ratio at t  =  5.7 s at ρ  =  0.7 for the simulation without radiation, in red on figure 26. The magenta circle corresponds to the $T_{\rm e}/T_{\rm i}$ ratio at t  =  5.7 s and at ρ  =  0.7 for the reference simulation. Note that JETTO is an iterative flux driven process while in QuaLiKiz stand-alone the gradients are kept fixed.

Zoom In Zoom Out Reset image size

On figures 27(a) and (b), ion and electron heat effective diffusivities both increase with the ratio $T_{\rm e}/T_{\rm i}$, as expected. It is coherent with the JETTO simulation, figure 26. Note that the slope of the increase of heat coefficients is quite stiff. As a consequence, a small modification of the $T_{\rm e}/T_{\rm i}$ ratio impacts significantly the turbulence. In this case, the removal of the radiation (i.e. the variation of heat diffusion between the red diamond and the magenta circle) caused an increase of 2.0 ${\rm m^2~s^{-1}}$ for the ion heat diffusion, 0.54 ${\rm m^2~s^{-1}}$ for the electron heat coefficient and 0.61 ${\rm m^2~s^{-1}}$ for the ion particle diffusion coefficient.

The removal of the radiation impacts the temperature and has a destabilizing effect. This indicates that a significant portion of the stabilization phenomenon occurs through the radiation. The next section focuses on the other mechanism susceptible to have a stabilizing effect: the combined effect of Z_eff stabilizing impact on ITG and increased collisionality.

The other mechanisms that could participate to the stabilization effect of W, are the combination of Z_eff stabilizing impact on ITG [42] and increased collisionality. Indeed, even if W is a trace impurity, it undergoes poloidal asymmetries and therefore can locally no longer be a trace species. Figure 28 shows the 2D poloidal cut of the W contribution to the Z_eff at t  =  6.5 s as an illustration. The W contribution to Z_eff remains similar at all times of the simulation.

Zoom In Zoom Out Reset image size

On the LFS, the poloidal asymmetries cause the W to contribute to Z_eff up to 0.5. In those zones, the W is no longer a trace species and can contribute to the ITG stability as well as modify the collisionality. Since the interchange ITG-TEM modes are ballooned on the low field side, where the W contribution to Z_eff is maximal it can indeed locally stabilize the turbulence [27, 42–44]. Moreover, W locally contributes to collisions, causing electrons to be untrapped and therefore lowering the TEM contribution. Both those effects caused by W can stabilize turbulence.

Using QuaLiKiz in stand-alone, the W concentration (calculated along $(R_{\rm max}-R_{\rm min})/2$ as the other input data) is increased and its impact on the effective diffusivities is illustrated by figure 29. The blue star corresponds to the zero W concentration for the simulation with Be only, in blue on figures 25. The magenta circle corresponds to the W concentration at t  =  6.5 s at ρ  =  0.7 for the reference simulation.

Zoom In Zoom Out Reset image size

On figures 29(a) and (b), ion and electron heat effective diffusivities remain unchanged until they reach a $\Delta Z_{\rm ef\,\!f}$ of ≃0.3. On figure 29(c), the main ion particle effective diffusivity follows a different trend, as it includes both the convective and diffusive contributions. The values of transport coefficients between the case without W and the W concentration from reference JETTO simulation (ie the difference between the blue star and the magenta circle) are of 3.1 ${\rm m^2~s^{-1}}$ for the ion heat diffusion, 0.97 ${\rm m^2~s^{-1}}$ for the electron heat coefficient and 0.62 ${\rm m^2~s^{-1}}$ for the ion particle diffusion coefficient. Therefore the contribution of the effective charge to the heat transport stabilization effect is 30% bigger compared to the contribution through the radiation impact illustrated on figure 27.

Overall, the W has a stabilizing effect on the turbulence. The effect occurs through radiation and the modification of temperature profiles, but also through the effective charge impact on ITG and collisionality.

JETTO settings:

Shot number	82722
Number of grid points	61
Start time (s)	5.5
End time (s)	7.1
Minimum timestep (s)	10−8
Maximum timestep (s)	10−3
Ion (1) mass	2
Current boundary condition (amps)	$1.67 \cdot 10^6$
Electron temperature boundary condition (eV)	102
Ion temperature boundary condition (eV)	102
Ion density boundary condition (${\rm cm}^{-3}$)	$1.5 \cdot 10^{13}$
Edge velocity boundary condition (cm s−1)	106
Grid resolution for QuaLiKiz	25

SANCO settings:

NEO settings:

QuaLiKiz settings:

$\rho_{\rm min}$	0.03
$\rho_{\rm max}$	0.97
Impact of $U_\parallel$, $\nabla U_\parallel$ and $E\times B$	Only for $\rho>0.5$
$k_{\theta}\rho$ range	ITG-TEM scales, ETG not taken into account here
Added diffusion coefficient	a Bohm diffusion of 0.1% of the particle diffusion coefficient,
To ensure numerical stability	1% a Bohm–GyroBohm heat diffusion coefficient [50]

ETB settings:

PENCIL settings: