The effect of impurity seeding into the closed divertor on plasma detachment in the HL-2A tokamak

Recent progress towards an increased understanding of detached divertor physics has been made with the highly closed divertor geometry in HL-2A. Non-intrinsic impurities were injected into the outer divertor chamber, and increased divertor neutral pressure and enhanced volumetric energy loss in the divertor were observed. Meanwhile the neutral pressure in the main chamber decreased slightly, and neutral compression between the divertor and main chamber increased greatly. This led to divertor detachment with a low upstream plasma line-averaged electron density ( ∼0.5nGW ). In the H-mode, slight degradation of the core confinement, characterized by a decrease in plasma stored energy and pedestal pressure and an increase in edge-localized mode frequency, was observed, but the H-mode was still sustained well with H98 > 1. Pedestal density fluctuation was increased during detachment, implying that the enhanced pedestal transport might be responsible for the degradation. During the divertor detachment phase, the impurities were well controlled in the divertor without strong radiation near the X-point region, and the main plasma density did not increase but decreased slightly; this could be a benefit of the highly closed divertor geometry. The experimental results suggest that a closed divertor geometry has the advantages of volumetric energy loss, gas pumping and impurity control in the divertor without significant effects on the plasma confinement, thus giving a wider operating window for divertor detachment.


Introduction
A low-temperature, low-heat-flux divertor operating regime called radiative detachment is required in ITER to reduce the heat loads on the plasma-facing components below 15 MW m −2 , while maintaining plasma confinement performance [1][2][3]. Non-intrinsic impurities have been used to enhance the divertor plasma radiation losses, and thus to achieve the divertor detachment regimes, such as in JET [4][5][6], ASDEX-Upgrade [7][8][9], DIII-D [10,11], Alcator C-Mod [12] and EAST [13,14]. However, a high-radiation zone near the X-point, i.e. multifaceted asymmetric radiation from the edge or MARFE, was often observed during the detached state, and this resulted in degradation of plasma confinement and even plasma disruption.
In principle, a higher degree of divertor closure could lower the impurity flux escaping from the divertor to the main chamber, which is beneficial for divertor detachment and preserves good plasma performance. Therefore, many devices have upgraded their divertors to increase the degree of closure by modifying the plate and baffle geometry, for example in ASDEX-Upgrade [15,16], JET [17,18], DIII-D [19,20], Alcator C-Mod [21] and JT-60U [22]. Here, the term 'closure' refers to the degree by which neutrals can escape the divertor region. The potential benefits from a closed divertor geometry are expected to be enhanced volumetric energy loss in the divertor, increased divertor neutral pressure for gas pumping and particle control, and reduced main chamber neutral pressure for reduced sputtering [23,24]. Dedicated experimental campaigns have been conducted in these devices. Neutral compression factors have progressively increased with closure and reached up to more than 100 with the Mark IIGB divertor in JET, which has a very high degree of closure [18]. Extended radiation patterns have been observed in ASDEX Upgrade Div II compared with Div I, and are in good agreement with modeling predictions [15]. The impacts of divertor geometry on the plasma characteristics have also been observed in DIII-D [19], JT-60U [22], Alcator C-mod [21] and JFT-2M [25]. In JET, DIII-D and JFT-2M, the increase in divertor closure has been accompanied by a decrease in main chamber neutral density, which would reduce the impurity contamination of the core plasma by reducing charge-exchange sputtering in the main chamber [23]. Furthermore, increased divertor neutral pressure allows energy diffusion, resulting in a reduction of the peak power flux. This leads to a reduction in the detachment upstream density threshold, thus broadening the operational parameter space with a detached divertor. More information about the effects of divertor geometry on detachment can be found in [23]. Recently, gas baffles have also been installed on TCV to decrease the coupling between the divertor and the plasma core, and the initial experimental results also confirm the predictions [26]. Therefore, a divertor geometry with a high degree of closure is attractive in divertor detachment operation.
One important criterion for a closed divertor geometry is to position the baffles at a distance corresponding to the field line at about two to three times λ n (the midplane scrape-off density width) from the separatrix [23]. In addition, compared with the horizontal targets, for the vertical targets the recycled neutrals are launched towards the separatrix, which becomes a region of preferential ionization, decreasing the temperature and increasing the density in its vicinity [2,23,27]. Meanwhile, for vertical targets, there is the possibility of adding a structure that prevents neutrals from escaping back into the bulk and scrape-off layer (SOL) plasma through the divertor region, increasing the divertor neutral pressure which favors divertor pumping, as reported in ASDEX Upgrade [28].
In HL-2A, the distance between baffles and targets at the entrance of the divertor is about 20 mm, which is almost the distance between the baffles and separatrix, d bs , as illustrated in the lower divertor configuration (see figure 1). This is about two to three times the scrape-off heat flux decay length, λ q [29], which is λ q ≈ 5 7 λ n according to the two-point model [27]. Therefore, the HL-2A closed divertor geometry meets the optimum requirements, as mentioned above, and has the potential to make uniquely relevant contributions to this research. Divertor detachment experiments have been carried out by injecting extrinsic impurities into the outer divertor chamber. As expected, high neutral compression between the divertor and main chamber was observed, which was induced not only by the increased divertor neutral pressure but also by the decreased neutral pressure in the main chamber. Divertor impurity radiation losses were also enhanced without observable impurity contamination in the main plasma. On this basis, divertor detachment of H-mode discharges has been achieved with high plasma confinement performance.
The rest of this paper is organized as follows. In section 2, a brief description of the HL-2A closed divertor geometry and related diagnostic systems is given. Characteristics of the divertor detachment in L-mode discharges are reported in section 3. The effects of the closed divertor geometry on the neutral pressure and radiation patterns are discussed in detail in section 4. Characteristics of detachment in the H-mode discharges and the impacts of edge localized modes (ELMs) on the detachment are presented in section 5. Finally, a summary and conclusions are given in section 6.

Divertor geometry and configurations
The HL-2A tokamak (major radius R = 1.65 m, minor radius a = 0.4 m) has a double-null divertor and a divertor geometry with a high degree of closure, where a closely spaced coil triplet with zero net current is used to produce the divertor configuration [29]. Additional ('multipole compensation') coils are added to cancel even the residual far-field of the two (symmetrically top and bottom) divertor triplets over the main plasma region, so that core flux surfaces are close to perfectly circular. The HL-2A tokamak is usually operated in the lower single-null (LSN) divertor configuration with the ion magnetic gradient drift towards the X-point, as shown in figure 1. The poloidal flux surfaces show that flux expansion is approximately a factor of two to three between the entrance to the divertor and the target plate region for the flux surfaces in the proximity of the separatrix. Furthermore, the conductance of the by-pass leaks is also reduced to a very low level with access to the H-mode. As a side effect, this physically highly closed divertor geometry allows for an efficient baffling of the neutral particles and a reduction of back-streaming into the main chamber. These effects offer the possibility to control the impurities with massive impurity gas injection in the divertor, and thus to realize plasma detachment operation without significant effects on the main plasma confinement.

Gas injection system and pumping system
The gas injection system for seed impurities (N 2 ) is shown in figure 1. The valves puff through dedicated cutouts of outer divertor baffle tiles, and the gas is delivered to the outer divertor chamber by a guide tube with a diameter of 1 cm. The amount of impurity gas sent to the divertor chamber is controlled by the fast-opening valves.
The pumping system on HL-2A is shown in figure 2, and has two components: two turbopumps and two cryopumps. There are arranged in toroidal symmetry, with a total pumping speed of S = 12 m 3 s −1 .

Relevant diagnostics
The HL-2A tokamak is equipped with a large number of plasma diagnostics, which provide the main parameters for the analysis of divertor detachment. Figure 2 gives an overview of the locations of some of the diagnostics used in this work.
A four-chord Michelson-type HCOOH (λ = 432.5 µm) laser interferometer (far-infrared, IR) is used to measure lineaveraged electron density below the midplane and along the horizontal lines of sight [30]. The outer midplane density profile is obtained up to 2 × 10 19 m −3 from the X-mode frequency-modulated continuous wave reflectometers [31]. The sampling rate of the density profile can be as high as 40 kHz, while the spatial resolution can be as small as ∼1 cm. The electron temperature profile is determined by an electron cyclotron emission (ECE) radiometer [32,33]. This system has a tunable local oscillator source, and can measure the second harmonic ECE frequency from 51 to 142 GHz. The spatial resolution is about 1.5 cm at a toroidal field of 1.3 T and the temporal resolution is about 1 µs. Density fluctuations are measured by Doppler reflectometry [34].
Two poloidal arrays of fast AXUV photodiode detectors are applied for estimating the total plasma radiated power along multiple chordal views in the main plasma chamber and lower outer divertor. Line radiation emitted by impurities in low-ionization states is measured by a passive visible spectroscopic diagnostic, which contains 20 lines of sight covering the X-point region and seven lines of sight viewing the lower outer divertor (green lines). This spectrograph is equipped with a movable holographic grating of 1800 rules mm -1 , blazed at 400 nm and 500 nm respectively, in order to cover the complete wavelength range of 185-860 nm [35]. Visible spectroscopy in the main chamber provides information on the injected impurities that enter the main plasma from the divertor chamber.
Divertor heat flux is calculated from the evolution of surface temperature on the target plates, measured by an IR camera [29]. The IR camera is optimized to have its best field of view covering the strike point zones, and to use a reduced array size with a 2 kHz sampling rate. Spatial resolution is as small as about 2 mm. Note that there is no inner target IR measurement because viewing is difficult with the highly closed divertor geometry. D α emission in the lower outer divertor is measured with a fast D α emission detector at a rate of 100 kHz [36]. Electron temperature in the divertor is estimated from the fixed Langmuir probes with a spatial resolution of 6 mm [37]. The fixed, flush-mounted Langmuir probes in the lower divertor are mostly triple probes. Single-probe operation is also available. The ion saturation current density, j sat , is derived with an effective probe area A eff . Neutral pressure measurements in the main chamber and divertor chamber are made by ionization gauges, separately, and the locations of these are shown in figure 2.

Detachment properties
The experiments were performed using a LSN diverted magnetic configuration with the ion magnetic field gradient drift towards the X-point. In order to understand the effects of divertor geometry on the detachment, Ohmic and L-mode regimes were chosen to avoid the effects of ELMs. The H-mode detachment and the effects of ELMs on that will be discussed in section 5.
Time evolutions for an example of divertor detachment with P NBI = 230 kW from 800 ms to 1300 ms are shown in figure 3. The electron temperature at the outer divertor target is measured with triple probes, and the errors in the temperature are estimated to be roughly ±50% from previous HL-2A experiments. The positions of the ion saturation current density and the electron temperature, shown in figures 3(d) and (e), are all around the outer strike point. At 800 ms, five nitrogen seeding pulses (t pulse = 10 ms, ∆t pulses = 60 ms) are injected in the outer divertor chamber with ∼4 × 10 18 particles per pulse. Following nitrogen injection, the intensity of NII (637.96 nm) line emission, which is approximately proportional to the impurity influx, in the outer divertor starts to increase with a small delay due to flowing gas. Due to the ionization of nitrogen atoms, the plasma density in the divertor increases and this results in an increase in the ion saturation current collected by the probes. Enhanced impurity radiation in the divertor dissipates most of the power flowing into the divertor before they reach the divertor targets. Furthermore, collisions between plasma and impurities also enhance the process. As a consequence, the electron temperature at the targets decreases. However, the electron pressure at 920 ms stays almost unchanged compared with that of early impurity injection at 850 ms (see figure 4(c)). This means that the divertor plasma is under a high-recycling regime before 940 ms [27]. With the increase in impurity influx, the characteristics of roll-over are shown in the ion saturation current, then there is a dramatic decrease in the ion saturation current at 940 ms. Meanwhile the electron temperature is further reduced to a lower level of 7 eV. These observations demonstrate that the plasma is detached from the divertor targets after 940 ms. In our experiments, the transition from a high-recycling regime to detachment seems to have occurred immediately without a continuous transition process. During the detachment phase, electron density in the main plasma remains at a very low level, ∼ 0.5n GW (n GW is the Greenwald density limit). In addition, the plasma radiation in the main plasma seems unchanged, hinting that the plasma radiation front is well controlled in the divertor without formation of MARFE at the Xpoint during the detachment state. These factors indicate that the main plasma is not affected by the impurities injected in the divertor, and so there is almost no degradation of the plasma confinement.
The profiles of ion saturation current, electron temperature and electron pressure measured by the Langmuir probe and heat flux measured by the IR camera on the outer divertor target for attached and detached conditions are compared in figure 4. The profiles are given within a time window of 20 ms. Unfortunately, there are no data for the SOL at the midplane because of a lack of diagnostics. At 850 ms, the plasma reaches steady state with neutral beam injection (NBI) heating, and the effects of injected impurities on the divertor conditions can be neglected at this moment (see figure 3). Therefore, the profiles at 850 ms are taken as the baseline for comparison. This approximation is reasonable because the profiles of upstream and divertor regions are comparable during a linear regime [27,[38][39][40].
In the high-recycling regime (920 ms in figure 4), because of enhanced impurity radiation losses the electron temperature is reduced substantially, but the reduction is only close to the strike point region. Meanwhile, due to the increased particle flux (Γ ∼ j sat ) the electron pressure stays almost constant, as does the divertor heat flux. The high-recycling conditions are sustained for about 70 ms (five to seven energy confinement times), then the outer divertor plasma undergoes rapid evolution to the detachment state (profiles at 1100 ms in figure 4), which is characterized by sharp reductions of the electron temperature, the electron pressure and the heat flux on the targets. During the detachment phase, the electron temperature is further decreased throughout most of outer divertor, while the particle flux is also reduced to very low levels. These result in a dramatic reduction of more than 90% in the electron pressure, and a reduction of 80% in the divertor peak heat flux.

Closed divertor geometry effects
The evolution of the neutral pressures in the main plasma and divertor during impurity seeding is shown in figure 5. With the onset of nitrogen seeding, the ionization gauge signal in the divertor increases rapidly, and then shows the trend of saturation. The pumped flux, Γ pump = n 0 S, can be derived from the neutral density, n 0 , in the pump chamber, and a pumping speed of S = 12 m −3 s −1 . Since neutral density is proportional to the pressure, the pumped flux can be estimated, and it increases by a factor of more than five with nitrogen seeding of the divertor. Moreover, the molecules that are formed by the incident ions from upstream are well confined in the closed divertor region. These result in a slight decrease in the neutral pressure in the main chamber, and also in the electron density in the main plasma. When the electron density is lower than the predicted value, the feedback system for controlling density is switched on (see figure 5(a) gas puff), and then the neutral pressure increases with gas puffing in the main chamber. It should be noted that although the value of neutral pressure in the main chamber is larger than that before the gas puffing, the electron density is not back to the value before gas puffing. With the impurity seeding, the divertor transitions into a detached state (after 940 ms), and the main plasma density remains at a low level.
A characteristic divertor performance measurement for divertor closure effects is neutral compression. This is defined by the ratio of the neutral density in the divertor and the main chamber, describing the ability of the divertor to enrich neutrals in the divertor for better pumping. In our experiments, the neutral compression varies in a range from 80 to 200. However, the ionization gauge in the divertor is located outside the outer divertor baffle (see figure 2), and the nitrogen gas is injected into the outer divertor, so the neutral density in the divertor should be much higher than the value obtained by the gauge.
The highly closed divertor geometry and divertor seeding result in high pumping, and thus will be good for preventing the escape of impurities directly from the divertor to the main chamber. Figure 6 shows the NII emission in the main and divertor chambers respectively. After nitrogen seeding, the NII emission increases greatly, and thus the energy deposited onto the divertor targets decreases; as a result, the heat flux drops immediately, indicating the start of detachment. The increase in NII emission in the divertor is much larger than in the main chamber, which again demonstrates that the impurities are well controlled in the divertor. In addition, the NII emission peak is localized in the divertor; it is far away from the Xpoint in figure 1 or figure 7 and does not move upwards during detachment phase, so there is less possibility of MARFE and the detachment state is stable in HL-2A.

Discussion
Experimental results in JET have demonstrated that the neutral compression increases with the degree of closure of the divertor, and a value of more than 100 has been achieved with a highly closed divertor (Mark IIGB) [18]. In HL-2A, the ionization gauge is installed outside the outer baffle, thus neutral compression is seriously underestimated. Figure 7 gives the SOLPS simulation results of neutral pressure distribution with the plasma parameters of shot #34493. It shows that the neutral pressure at the location of the ionization gauge is about 0.5 Pa, consistent with the experimental results. Meanwhile, the neutral pressure in the divertor is much higher than that at the location of the ionization gauges. In this case, the neutral compression can reach up to 1000 in the highly closed divertor geometry of HL-2A, which is in agreement with the expected beneficial effects of the a highly closed divertor geometry.
Although increased neutral compression has been observed in many devices after increasing the degree of closure of the divertor, the increase in neutral compression is caused only by the increase in divertor pressure in Alcator C-mod [21] and ASDEX Upgrade [41]; the decrease of the main chamber pressure also contributes to the increase of neutral compression in JET [18,42] and DIII-D [43]. As mentioned above, a slight decrease in the main chamber pressure is observed following impurity injection, but it is affected by gas puffing in the main chamber in this case. Figure 8 shows a clear variation of the neutral pressures, in which D 2 is injected into main plasmas by the supersonic molecular beam injection (SMBI) technique [44]. It should be note that there is no gas or impurity injection in the divertor region. Following SMBI, the neutral pressure in the divertor and neutral compression are greatly increased, thus resulting in high pumping. Consequently, a clear reduction in the main chamber pressure is observed, although D 2 is injected into the main chamber. After SMBI, the divertor neutral pressure decreases while the neutral pressure in the main chamber increases to its previous value. One reason for this may be the highly closed divertor geometry, which has a very low conductance for neutral backflow into the main chamber.  Another possible reason is the enhanced pumping with the increased neutral compression. Thereby, the HL-2A results are consistent with the results of JET [18,42] and DIII-D [43]. Figure 9 shows the ratio of NII radiation power between the divertor and main plasma to demonstrate impurity screening. The NII radiation in the divertor is larger by a factor of about 40 than that in the main plasma. Meanwhile, the impurity radiation peak is localized in the divertor without moving to the X-point during impurity injection (see figure 6), even during the detachment phase. Compared with the results for ASDEX Upgrade [15,16], JET [18] and DIII-D [43], this demonstrates that a highly closed divertor geometry can prevent the impurities from escaping from the divertor to the main chamber, and this is beneficial for the control of impurity radiation patterns and high performance control.

Detached H-mode plasmas
Divertor detachment experiments during H-mode discharges were also carried out by injecting non-intrinsic impurities into the outer divertor chamber.

Characteristics
H-mode discharges have been achieved with NBI heating, electron cyclotron resonance heating (ECRH) and lower hybrid current drive (LHCD) in HL-2A [45][46][47]. The main parameters for H-mode plasmas are plasma current But compared with the L-mode, the amount of impurity gas is enhanced because of the high heat power. In addition, the heat flux deposited on the outer divertor targets is also reduced dramatically, as shown in figure 11, especially near the separatrix. This is well correlated with the decrease in the electron temperature. In the private flux region and far SOL region, the heat flux shows a slight increase, which should be caused by the enhanced radial energy transport in the divertor as already discussed in ASDEX Upgrade [15]. Accompanied by impurity injection, the ELM frequency increases from about 200 Hz to 400 Hz for shot #36898 and ELM size decreases (see figures 10(a) and (e)), similar to other experimental results [7,14,42,48].

Pedestal, confinement and ELM effects
Following nitrogen injection, the plasma stored energy decreases slightly as does the H 98 factor, as shown in figure 12. As the other parameters, such as plasma current, injected power of NBI and LHCD, do not change, the confinement degradation should be caused by the impurity contamination.
After impurity seeding, the NII line emission is observed immediately in the main plasma by visible spectroscopy, and intensity shows the tendency to increase slightly (see figure 12(g)). Because the highly closed divertor geometry has a very low conductance for neutral backflow into the main  chamber, the impurity ions in the main plasma should be due to the thermal gradient force [27]. The ELM causes a sudden crash of the transport barrier with transient releases of energy and particles into the SOL and eventually towards the divertor. Therefore, the detached plasmas are burned through due to the energy flux ejected by ELMs, and there is a broadening of the SOL during ELM events, despite the enhanced energy dissipation by impurity radiation in the divertor. As a consequence, the thermal force pushes impurity ions away from divertor and into the main plasma, and the region of impurity flow reversal is concentrated close to the separatrix where the thermal force is largest. However, the intensity of NII emission in the outer divertor is about 30 times that in the main plasma, indicating that the impurity content in the main plasma is much less than that in the divertor, despite the ELM effects. In addition, the total plasma radiation in the main plasma only increases a little after impurity injection, as shown in figure 12(e). This is in agreement with the low impurity flux in the main plasma. Although the core impurity contamination gives rise to some extent of confinement degradation during the detached state, the H 98 factor is still larger than 1.
Similar to the L-mode, the line-averaged electron density decreases slightly after impurity seeding (see figure 12(c)), which is also observed in TCV [26]. As expected, a closed divertor geometry increases divertor neutral pressure for gas pumping. When impurity gas is injected into the outer divertor chamber, the neutral pressure in the divertor is greatly enhanced, resulting in an increase in pumping. This leads to a decrease in the neutral density in the main chamber, and this would result in a decrease in the main plasma electron density.
The slight decrease of H 98 in confinement indicates that pedestal transport is changing. During the detached state (1118 ms), the pedestal density decreases compared with that of the attached state (980 ms), while the pedestal temperature seems to be unaffected, as shown in figure 13. This means that a reduction in electron pressure occurs around the pedestal region, leading to an increase in ELM frequency and a decrease in ELM size (see figure 10). With an open divertor geometry, the electron density usually increases after detachment [7]. This implies that a closed divertor could prevent neutral particles moving back into the main plasma, and might be beneficial for particle control compared with an open divertor.
As discussed above, some impurity ions flow into the main plasma due to the thermal force. As reported in [49][50][51], these ions reduce the velocity shear rate by changing the toroidal velocity term, and the reduction of the E × B velocity shear rate enhances the pedestal turbulence, possibly through turbulence spectral shift. Figure 14 shows the amplitude spectrogram of Doppler reflectometry, which measures plasma turbulence. It shows that after impurity injection at 1000 ms the turbulence intensities near the pedestal  region become stronger. This enhances the density transport as explained in the experimental reports [49][50][51], and this should be responsible for the decrease in pedestal density.

Summary and conclusion
The HL-2A tokamak has a unique highly closed divertor geometry. As expected, the highly closed divertor results in: intense radiation loss in the divertor to reduce the energy deposited onto targets; a high divertor neutral pressure in the divertor for gas pumping and particle control; a reduced main chamber neutral pressure for reduced sputtering; better impurity screening for a high performance capability. Neutral compression between the divertor and main chamber can reach to more than 100, which is seriously underestimated due to the location of the ionization gauge. Furthermore, a low level of impurity flux in the main plasma is observed during impurity injection in the divertor, which is consistent with predictions [26].
Divertor detachments are obtained both in L-and H-modes by injecting impurities into the divertor to enhance radiation losses. Sharp reductions in the electron temperature are observed at the start of detachment. During the detachment phase, the impurities are well controlled in the divertor, and the ratio of impurity flux between the divertor and main plasma is more than 50. But a small number of impurities escape from the divertor flowing into the main plasma due to the thermal force, especially during ELMs. These impurity ions drive turbulence around the pedestal region, resulting in a reduction in pedestal pressure. As a result, an increase in ELM frequency and a decrease in ELM size are observed. However, the plasma remains in the H-mode regime with H 98 > 1. Pronounced detachment is achieved with a very low plasma density (∼ 0.5n GW ). Furthermore, the divertor detachment regime is stable, without any formation of intense radiation around the X-point (MARFE), which is also confirmed by the tangential visible CCD camera. Therefore, the experimental results demonstrate that the very closed divertor geometry can prevent the direct escape of neutrals from the divertor to the main chamber, and mainly enhance the divertor radiation losses and control the impurities, thus benefitting gas pumping, particle control, divertor detachment and performance compatibility.