Achievement of radiative feedback control for long-pulse operation on EAST

The excessively high heat load on divertor target plates is one of the most critical issues for magnetically confined fusion devices, such as tokamaks, stellarators and future fusion reactors. For an ITER divertor, the tolerable steady-state peak heat flux is about 5–10 MW m⁻² [1, 2], which is below the unmitigated heat load for an attached divertor condition [1]. Therefore, active control technology of the heat flux must be employed to protect the divertor targets from overheating and prevent impurity production. The most promising means for steady-state heat flux control is impurity seeding during the plasma discharge, especially for superconducting tokamaks like EAST, ITER and fusion reactors. Seeded impurities can convert a large fraction of the thermal energy into radiated power, and thus reduce the peak heat flux and total power incident on the divertor target plates. So far, many tokamaks have explored and implemented this method of controlling the radiation in order to control the divertor heat flux, such as ASDEX-Upgrade [3–5], JET [6, 7], DIII-D [8, 9], JT-60U [10] and TEXTOR-94 [11]. The active feedback control of a radiative divertor has advanced significantly in recent years on ASDEX-Upgrade, using two separate impurity species from upstream and divertor regions simultaneously [12]. However, all the radiative feedback control experiments are carried out in short pulse (<10 s) discharges, since the above tokamaks are not superconducting machines. The extension of radiative feedback control means a long-pulse superconducting tokamak is of great urgency for the forthcoming ITER.

The experimental advanced superconducting tokamak (EAST) is the first tokamak with fully superconducting toroidal and poloidal field coils. It has a major radius of R  =  1.9 m and a minor radius of a  =  0.5 m. EAST is designed for long-pulse steady-state high performance plasma operation over 1000 s [13], which provides a good platform for the active feedback control of radiation in long-pulse operations. Currently, in long-pulse H-mode discharges of over 60 s on EAST, the steady-state heat flux on the divertor targets is about 3–4 MW m⁻², with a heating power of about 5 MW [14]. Excessive high heat flux on the target plates is also a critical issue for near-term operation on EAST, since the capability of total auxiliary heating power has been expanded to more than 20 MW [15]. This paper will present an approach to set up an active feedback control system for radiated power on EAST to accommodate long-pulse operation with an ITER-like tungsten divertor.

The specific technical implementation of the radiative feedback control system and preliminary experimental results will be shown. The remainder of this paper is organized as follows: section 2 illustrates the implementation of the radiative feedback control system and algorithm in detail. Section 3 includes three main parts: (1) summarizing the experimental conditions and main diagnostics for this research; (2) estimating the capability of the control system and the confinement performance of the plasma; and (3) analyzing the behavior of particle flux and temperature on the divertor target plates. Finally, section 4 gives a brief summary and discusses future improvements for the feedback control system.

Collision radiation theory and the coronal model [16] show that impurity particles injected during a plasma discharge will undergo complicated radiation processes which will reduce both the heat and ion flux before the particles strike the divertor target plates. The main radiation loss processes in plasma are Bremsstrahlung radiation, line radiation and recombination radiation. Based on these observations, the basic means of plasma radiative feedback control is seeding an impurity gas when the measured radiated power is lower than a set value, i.e. the heat flux to the files is high, which is determined to be safe for particular modes of tokamak operation. This requires the radiated power to be exactly calculated in real time, and thus the impurity seeding quantity can be adjusted automatically via a controller—i.e. a proportional-integral-differential (PID) controller—in our experiments.

2.1. Actuators of the control system

2.2. Logic of radiative feedback control algorithm

The logic of the overall control system is divided into two parts: (1) radiative diagnostic data processing in real time and (2) radiated power feedback control. In EAST, there are 64 absolute extreme ultraviolet (AXUV) [21] channels covering the poloidal extent of the vacuum chamber. A data processing subsystem is configured with a DTACQ 196 digitizer and reflective memory (RFM) for data transmission with PCS, as shown in figure 1. The data processing subsystem starts to work when a new shot number is received through socket communication with the PCS. Execution is synchronized to the PCS with the same control cycle to within 100 µs. During each cycle, the radiated power is calculated using AXUV signals and plasma boundary data, and the result is sent back to the PCS. Detailed information about the power calculation is described in section 2.3.

Zoom In Zoom Out Reset image size

Radiated power feedback control is realized in the PCS, including a flexible interface setting and control logic design. As shown in figure 1, when the radiated power calculation result is obtained from the data processing subsystem, a control difference/offset error is generated by comparing the value to a reference target, then the PID controller is applied to the error to output the feedback calculation command. When the radiated power is lower than the target, the gas puffing commands will be sent to the gas puffing system, which includes the divertor PE valve and SMBI through the RFM. The use of impurity gas seeding from the feedforward divertor actuator is essential to initially elevate the radiated power. This step makes the radiated power increase significantly so that the feedback SMBI actuator with the impurity can then easily follow the radiated power further up to the set reference target value and maintain it stably during the entire feedback control phase.

2.3. Radiated power calculation and impurity species selection

In EAST, the AXUV photodiode diagnostic can be used for the absolute total radiated power measurement due to its wide photon response from 7 eV to 6 keV [21, 22]. Four 16-channel AXUV arrays were installed in the EAST horizontal diagnostic port, which can view nearly the whole poloidal plasma cross-section with a spatial resolution of about 3–5 cm. The poloidal layout of the AXUV viewing chords is shown in the lower right part of figure 1. For fast real-time calculation, the total radiated power of the bulk plasma is calculated by directly summing the line-integrated radiation intensity according to the following equation (1):

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{P}_{{\rm rad}}}=\sum\limits_{i}{2\pi {{R}_{i}}\Delta {{r}_{i}}{{C}_{i}}{{X}_{i}}}\nonumber \end{align} \tag{ 1 }$$

where ${{X}_{i}}$ is the single-channel raw voltage signal of the AXUV diagnostic output. ${{R}_{i}}$ and $\Delta {{r}_{i}}$ are the large radii of each viewing chord and the distance between two adjacent viewing chords separately, which are calculated in real time from the RTEFIT plasma equilibrium [23]. The coefficient ${{C}_{i}}$ is the calibration constant which is specific to each channel, and has been calibrated for each channel. This is mainly determined by three parameters: (1) the parameters of the detectors of the AXUV array, (2) the parameters of the amplifiers in the AXUV circuit, and (3) the geometry layout of the AXUV channels. A specialized computing server for calculating the real-time total radiated power of bulk plasma with equation (1) has been equipped and utilized as shown in figure 1. It should be mentioned that not all the 64 horizontal channels are needed for this calculation, and the overlapping channels are excluded, because they will lead to a double calculation of radiated power in some regions.

The main choices for the impurity gas species are low or medium Z elements such as nitrogen (N₂), neon (Ne), CD₄ and argon (Ar). In DIII-D radiative divertor research, D₂ gas was also used to enhance the radiative processes [24]. A few high Z impurity gasses like krypton (Kr) are also currently being investigated [25, 26]. The results on ALCATOR C-Mod [27], ASDEX-Upgrade [28] and JET [29] illustrate that nitrogen is well suited and even better than other impurities, like Ne and Ar, at simultaneously controlling the divertor heat load and maintaining good confinement due to the N₂ radiating power concentrated in the edge region. However, N₂ is not a suitable choice for EAST due to the application of lithium (Li) for wall conditioning [13], because N₂ can easily interact with Li to produce Li₃N easily, which seriously exhausts the Li coating. The generated Li₃N can also interact with H₂O to produce NH₃ to erode some metal materials. In previous EAST experiments, Ar has been chosen as the impurity gas to be injected into the divertor region, which yielded an effective reduction in heat and particle fluxes towards the divertor target plates, and also led to significant energy loss in the core region [30–32]. Compared with Ar, Ne is much milder due to its lower core radiation. The behavior of the Ne injection on EAST has previously been modeled [33], and shows that localized Ne puffing is indeed effective at reducing the divertor heat flux, so Ne is a good choice for the active control of radiated power. In EAST's present operation regime, Ne impurity seeding favors the edge and SOL radiation. Therefore, Ne seeding experiments have been carried out in the recent EAST experimental campaign and the Ne impurity was chosen as the primary impurity for the radiative feedback control experiments reported in this paper.

In this paper, all plasma discharges are carried out in both L-mode and H-mode regimes with the plasma current ${{I}_{{\rm p}}}$  =  0.4 MA, the safety factor ${{q}_{95}}$ ~ 4.6, and the toroidal field ${{B}_{{\rm t}}}$ clockwise when viewed from the top. In H-mode, the auxiliary heating systems used are low hybrid waves (LHW) of 4.6 GHz with ${{P}_{{\rm LHW}}}$  =  1.5 MW and neutral beam injection (NBI) from two sources with ${{P}_{{\rm NBI}}}$  =  1.9 MW. In L-mode discharges, the heating power is only supplied by the 4.6 GHz LHW at ${{P}_{{\rm LHW}}}$  =  1.5 MW. The line-averaged electron density is between $3.0\times {{10}^{19}}$ m⁻³ and $4.0\times {{10}^{19}}$ m⁻³. Note that all the experimental results reported here are obtained in an upper single-null (USN) configuration with an ITER-like tungsten divertor, and the Li coating [34] conditions are the same in these discharges.

The particle flux on the divertor target plates is measured by the divertor Langmuir probe (DivLP) system, which consists of 89 sets of triple probes in the upper tungsten and lower graphite divertors. The DivLP system is embedded in both the inner and outer divertor target plates. For the upper DivLPs, the poloidal spatial resolution is 1.2–1.8 cm, with the temporal resolution being 0.02 ms (50 kHz) [35]. The temperature of the divertor target plates is measured by infrared cameras (IR) located in the horizontal port, which can view both the upper and lower divertor plates and provide infrared images with a 320  ×  240 pixel size [36]. Impurity line emissions in the core plasma are measured by a fast-time-response extreme ultraviolet (EUV) spectrometer working at 2–50 nm [37]. It can monitor various impurity species including C, Ne, tungsten (W), etc. The ion temperature of the core plasma is measured by charge exchange recombination spectroscopy (CXRS). The core CXRS consists of 30 channels and covers a radial range of 1.55–2.33 m with a radial spatial resolution of 0.5–3.0 cm [38], and the edge CXRS consists of 32 channels with a radial spatial resolution of about 7 mm [39]. The electron temperature of the plasma is measured by the Thomson scattering (TS) system, which has a spatial resolution of 1–2 cm [40].

The active control system of the radiated power is effective for both L-mode and H-mode operations in EAST. In all the control experiments, the plasma loop voltage, ${{V}_{{\rm loop}}}$, is well controlled at zero, so it is easy to elongate the plasma pulse length. In the H-mode experiment, the preset control level of the total radiated power in the bulk plasma was gradually elevated from 0.6 to 1.2 MW over six shots, i.e. ${{f}_{{\rm rad}}}$  =  18%–36% with a background radiated power (i.e. without impurity injection) of about 0.5 MW. In L-mode operation, the control range was from 0.3 to 0.45 MW (${{f}_{{\rm rad}}}$  =  20%–30%), while the background radiated power was about 0.17 MW. In both the H-mode and L-mode, the radiated power was successfully maintained at the preset level during the active feedback control phase. Figure 2 shows the control results for a series of sequential H-mode discharges (2(b)–(d)) and H-mode discharge shot #71022 without radiative feedback control (2(a)). It should be mentioned that in #71022, one of the NBI sources unexpectedly stopped working at about 5.7 s, which made the level of stored energy decrease abruptly. The feedforward PE valves in the divertor region raise the total radiated power of the bulk plasma in a large, coarse adjustment early in the discharge (at 3.0 s), while short Ne pulses injected by SMBI (from 4.0 s to 7.5 s) are used to reach and maintain the desired radiated power with finer adjustments. The radiated power is influenced by many physical and chemical processes, like bursts of heavy impurity radiation, so the actual radiated power oscillates slightly around the set level as shown in figure 2. It should be noted that the burst of radiation at t  =  4.66 s in shot #71021 is contributed by Cu ions, which may be transported into the plasma after being produced at the antenna surface through a physical sputtering process. Table 1 summarizes the maximum error range of the radiative control system for the different radiated power set levels. It should be mentioned that the results in table 1 exclude the abnormal values caused by heavy impurity bursts, which are not part of the control error. As the desired level of radiated power increases, the relative control error decreases.

Table 1. The range of maximum control error for different radiation power set level.

Shot no.	${{\boldsymbol{P}}_{{\bf rad,target}}}$(MW)	Control error
71018	0.6	[−24%, +14%]
71019	0.8	[−20%, +10%]
71021	1.0	[−18%, +6%]
71020	1.2	[−13%, +4%]

Zoom In Zoom Out Reset image size

Figure 3 shows the EUV spectra at 20–120 Å for two time slices before and after a group of short Ne feedback pulses in shot #71021, which are also indicated in figure 2(d) with vertical dotted lines. Line emissions from C, W and Ne ions are indicated in the spectra, i.e. CVI 33.73 Å and its 2nd order, tungsten unresolved transition array (W-UTA) at 40–75 Å composed of W²⁷⁺–W⁴⁵⁺, NeVII 106.19 Å, NeVIII 88.09 Å, NeVIII98.26 Å and NeVIII 103.09 Å. It can be seen that after Ne injection, the intensity of the Ne VII and Ne VIII lines obviously increased, and the intensity of W-UTA also slightly increased, which indicates that additional erosion of W was produced by Ne injection.

Zoom In Zoom Out Reset image size

Plasma performance is influenced slightly by the operation of radiative feedback control. The plasma-stored energies, ${{W}_{{\rm mhd}}}$ and ${{H}_{98}}$, are important parameters concerning plasma performance. The reduction of ${{W}_{{\rm mhd}}}$ and ${{H}_{98}}$, caused by Ne impurity injection during the feedback control phase, is minimal and well within the acceptable range. For preset levels with different radiated power in H-mode operations, the range of ${{W}_{{\rm mhd}}}$ variation is 7%–13%. Figure 2 also shows the time traces of ${{W}_{{\rm mhd}}}$ and ${{H}_{98}}$ during the radiative feedback control phase. As an example, during shot #71021 the original radiated power without impurity seeding does not exceed 0.56 MW. When the radiated power reaches 1 MW and is maintained stably by radiative feedback control, the plasma remains in H-mode and the plasma-stored energy decreases slightly by 13 kJ, which is about 8% of 165 kJ without impurity seeding. Similar to ${{W}_{{\rm mhd}}}$, the time tendency of ${{H}_{98}}$ during a radiative feedback control phase is kept at a high level with just a slight reduction. Thus, ${{W}_{{\rm mhd}}}$ and ${{H}_{98}}$ can corroborate each other, which further proves that radiative feedback control has no serious negative impact on plasma performance. The impurity seeding may lead to a drop of temperature in the main plasma. Since the duration of the SMBI-Ne pulses is shorter than 3 ms, the influence of each single impurity pulse on the core electron/ion temperature does not last for a long time—but the cumulative effect in a control phase may be significant. Figures 4(a) and (b) show the profiles of the electron temperature (${{T}_{{\rm e}}}$) measured by TS and the profile of the ion temperature (${{T}_{{\rm i}}}$) measured by CXRS, and the corresponding fitted curves based on the diagnostic data are also shown respectively. At different time slices of the radiative feedback control phase, shown in figure 4(c) as vertical dotted lines, the influence on the ${{T}_{{\rm i}}}$ profile is not very significant, and the maximum drop is about 14%. The ${{T}_{{\rm e}}}$ profile drops after the feedforward Ne injection, but during the feedback control phase, the ${{T}_{{\rm e}}}$ profile was nearly unchanged and slowly rose at 6.9 s. The reduction of ${{T}_{{\rm e}}}$ and ${{T}_{{\rm i}}}$ may result from the radiated loss of Ne and the increase of electron density, ${{n}_{{\rm e}}}$ (not shown), due to Ne seeding. However, the current Ne injecting level does not make n_e increase significantly and produce unnoticeable impact on ${{n}_{{\rm e}}}$ profile. As mentioned above, the radiative control experiments on EAST were operated under the same Li wall condition. Li coating is an effective means to reduce intrinsic impurity radiation and suppress impurity flow into the core plasma [41, 42]. Thus, when implementing radiative feedback control without Li coating, the intrinsic radiation of the plasma will be higher, and it may need a lower Ne gas injection volume to achieve the same radiated power level under Li coating conditions.

Zoom In Zoom Out Reset image size

More importantly, effective reduction of the divertor target temperature with the active feedback control of radiated power was achieved in the dedicated experiment. The ultimate objective of radiative feedback control is to prevent the divertor target plates from overheating by reducing divertor heat flux and eventually lowering the actual surface temperature of the target plates. Energy conservation in tokamak operations can be described by equation (2) [41]:

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{P}_{{\rm heat}}}={{P}_{{\rm rad}}}+{{P}_{{\rm load}}}+\frac{{\rm d}{{W}_{{\rm mhd}}}}{{\rm d}t}\nonumber \end{align} \tag{ 2 }$$

where ${{P}_{{\rm heat}}}$ is the input heating power, ${{P}_{{\rm rad}}}$ is radiated output power, ${{P}_{{\rm load}}}$ is the heat load on the divertor target plates and other PFCs, and ${\rm d}{{W}_{{\rm mhd}}}/{\rm d}t$ is the derivative of plasma-stored energy with respect to time. With the same heating power, active feedback control makes the radiated power increase, while keeping the stored energy nearly constant. Therefore, the heat load on the divertor target plates should be reduced during the control phase. Unfortunately, heat flux data from the upper divertor on EAST is not available, but the divertor temperature measured by IR camera and the particle flux measured by the DivLPs still demonstrates the variation of divertor heat flux in the radiative control phase. When radiative feedback control is operated during H-mode discharge, the temperature of the target plates measured by the IR cameras is obviously reduced and maintained at a lower level compared to that without Ne gas injection. Figure 5(a) shows the temperature of the upper outer divertor target plate in shot #71021 measured by IR camera. When Ne gas was injected from the PE valves in the divertor region, the temperature started to drop. For the entire duration of the feedback control phase, the temperature of the divertor target plates was maintained at a lower level, and started to increase when the feedback was switched off. In the strike point region of the outer target plates, the temperature descends around 250  K–300 K during the feedback control phase. The divertor temperature indicates the accumulation of the divertor heat flux. The reduction of the target temperature suggests that the heat flux incident is well controlled on the divertor target.

Zoom In Zoom Out Reset image size

Since the heat flux is the major concern here, it should also be noted that part of the reason for the heat flux reduction is the reduction of particle flux on the target plates. Figure 6 shows the time traces of the peak particle flux (${{\Gamma }_{{\rm peak}}}$) at the inner and outer divertor target plates for different discharge conditions measured by the EAST divertor Langmuir probe system. In the L-mode discharge, shot #70949, the descending rate of ${{\Gamma }_{{\rm peak}}}$ is about $65.8 \%$ for the upper inner divertor and $45.8 \%$ for the upper outer divertor after the feedforward Ne injection, as shown in figure 6(a). For the USN discharge with $B\times \nabla B\downarrow$ in shot #70949, the feedforward Ne was from the lower outer divertor PE valve, and mainly transported to the upper inner divertor with SOL flow [43], which can partially explain why there is a preferential impact on the upper inner target ${{\Gamma }_{{\rm peak}}}$ compared to the upper outer target. The Ne gas puffed by the feedforward valves caused the peak particle flux to reduce significantly; it was then gradually recovered, but the overall amplitude of the ${{\Gamma }_{{\rm peak}}}$ stayed at a low level later. In H-mode USN discharges with $B\times \nabla B\downarrow$, #71021, the peak particle flux did not decrease after feedforward Ne injection. The feedforward Ne gas from the PE valve at the upper inner divertor was injected into a private flux region (PFR). Based on the drift direction, Ne gas was transported from the inner target to the outer target in the PFR, but the amount of particle release from the core plasma increased in H-mode, which seriously weakened the Ne injecting effect. Thus, the peak particle flux was not sharply reduced after a feedforward injection like an L-mode situation. The Ne injection from the SMBI has nearly no influence on the divertor peak particle flux either in L-mode or H-mode, because the injecting volume of Ne gas at any time from the SMBI is small. It can also be seen that the Ne gas pulsed during the control phase has almost no impact on the amplitude or frequency of the ELMs in the H-mode discharge (figure 6(b)), which also suggests that the quantity of Ne injected by the SMBI system is limited. Both ELM mitigation and suppression have been observed in EAST when the pulse width of the SMBI-Ne increases [44].

Zoom In Zoom Out Reset image size

The active feedback control of radiated power with neon impurity seeding has progressed significantly in EAST operations experimentally. Fast impurity seeding was realized by employing SMBI. For long pulse discharges, the operating state of this feedback control system is stable, with radiated power being actively adjusted over a wide range to ensure good controllability. The total duration of radiative control (feedforward and feedback) is 4.5 s in a typical plasma discharge of 10 s. In H-mode discharges with an auxiliary heating power of 3.3 MW, an adjustable range was successfully achieved from 0.6 to 1.2 MW, i.e. ${{f}_{{\rm rad}}}$  =  18%–36%. In L-mode discharges with a heating power of 1.5 MW, the radiated adjustable power range is from 0.3 to 0.45 MW, i.e. ${{f}_{{\rm rad}}}$  =  20%–30%. The maximum relative control error decreases from 24% with the desired radiated power of 0.6 MW to 13% with the desired radiated power of 1.2 MW in long-pulse H-mode experiments. The ${{f}_{{\rm rad}}}$ of 36% is not the upper limitation of our radiative control system, because for ITER, it may need an ${{f}_{{\rm rad}}}$ of up to 60% or more to prevent the divertor target overheating [45]. For the current feedback control experiments on EAST, because the SMBI is a gas puffing device with a high injecting efficiency and is close to the core plasma, excessive neon seeding from SMBI will cause the serious degradation of plasma confinement. Thus, we have to be cautious about increasing ${{f}_{{\rm rad}}}$. In this controllable range, we realized the control of radiated power without a serious negative impact on the plasma performance. In the radiated power range within 0.6–1.0 MW under H-mode discharges, the plasma-stored energy loss is 7%–13%, and ${{H}_{98}}$ is also kept at a high level with a slight reduction. So, the degradation of plasma confinement caused by neon seeding is tolerable. In the future, the ${{f}_{{\rm rad}}}$ will be improved further, and further details of the EAST divertor system upgrade will be stated in the next paragraph. During the active feedback control phase, the temperature of the divertor target plate was effectively decreased, with the reduction being about 250 K–300 K. The Ne injection from SMBI had no influence on the divertor peak particle flux resulting from the very small injection volume. Despite the current radiative control producing additional W erosion, our final aim is to realize controllable partial/complete detachment by a radiative feedback control method, reducing the divertor temperature below 5 eV, which will effectively suppress the erosion and sputtering of W and prolong the lifetime of the tungsten divertor.

In the future, the active feedback control system will be extended to even longer discharges, higher radiation fraction and more advanced plasma configurations like quasi-snowflake (QSF) divertor operation. The lower divertor in EAST will be upgraded from graphite to tungsten. Along with this, one of the most important upgrades will be the divertor impurity gas puffing system. The fast gas puffing system, i.e. SMBI, which has shown powerful capability for real-time divertor radiation feedback control, will be utilized in the divertor region. Using this impurity seeding injection upgrade, fast impurity puffing far from the core plasma is hoped to further reduce the impact on plasma confinement, in comparison to the present LFS mid-plane injection. More importantly, this added capability will enhance divertor radiative feedback control handling. By shifting the total radiated power distribution of bulk plasma with feedback control from the core to the divertor volume, integrated control of the heat flux and core plasma performance in long-pulse operations will be exploited.

This work was supported by the National Magnetic Confinement Fusion Science Program of China under grant nos. 2014GB103000, 2015GB101000, the National Natural Science Foundation of China under grant nos. 11575245, 11575236, 11575242 and 11625524, and the National Key Research and Development Program of China under grant no. 2017YFA0402500.