A review of supersonic molecular beam injection for plasma fueling and physical studies in magnetic fusion devices

Abstract

In magnetic confinement fusion devices, supersonic molecular beam injection (SMBI) is commonly used as a fueling method, which has also performed well in physical studies since it was first proposed on the HL-1 tokamak by Southwestern Institute of Physics. This study presents the development of the SMBI technique since its first use in fusion experimental devices and reviews the progress on the investigation of plasma physics using the SMBI. In addition, this study further discusses the potential applications of the SMBI technique on future fusion devices.

1 Introduction

Plasma fueling and its density control are crucial issues for magnetic confinement fusion reactors such as ITER (Loarte et al. 2007). A high plasma density is favorable for high-performance operations. On the one hand, for the optimal plasma temperature, a high plasma density can help achieve a high fusion triple product $n\tau T$, where $n,\tau$, and $T$ are the plasma density, energy confinement time, and ion temperature, respectively. On the other hand, a high density is required for future ITER operation with high dissipative divertor conditions to reduce the power reaching the targets. However, plasma fueling in large tokamaks faces several challenges. First, the design concern of the fueling system jointly depends on the core and divertor plasma statuses, which are difficult to accurately predict. Consequently, these systems must be designed with reasonable margins to allow for uncertainties in plasma conditions. Second, operational scenarios with high plasma density and temperature require a fueling system to achieve deeper penetration for high fueling efficiency. Finally, the complex working conditions of the fusion reactors, including the high magnetic field and high neutron flux, should be considered in the design of the fueling system. Presently, in addition to gas puffing (GP) and pellet injection (PI) (Baylor et al. 1878), supersonic molecular beam injection (SMBI) is envisaged as an alternative for fueling and controlling particles in fusion plasma devices. SMBI was first used as a plasma fueling method in the HL-1 tokamak, and then it was shown to have a high fueling efficiency on HL-1 M (Yao et al. 2004). Its high fueling efficiency was also demonstrated in several tokamaks, such as HL-2A (Yao et al. 2007, 2010), Tore Supra (Pegourie et al. 2003), and EAST (Zheng et al. 2013). Thus, it has been widely implemented in many notable fusion research devices, such as ASDEX Upgrade (Lang et al. 2005), NSTX (Soukhanovskii et al. 2004), JT-60U (Takenaga et al. 2009) KSTAR (Kim et al. 2013), LHD (Murakami et al. 2012), Heliotron J (Mizuuchi et al. 2010), J-TEXT (Xiao et al. 2014a), W7-AS (Yao and Baldzuhn 1933), ADITYA-U (Tanna et al. 2019), and GAMMA 10 (Islam et al. 2016). Owing to its flexibility and fast response, the SMBI technique has also been well developed for density feedback control on HL-2A (Chen et al. 2016) and EAST (Zheng et al. 2013). The experimental results show that density feedback control with SMBI performs much better in both fast and accurate control of the density than the normal gas puffing.

Considering the installation location and ports, the SMBI system is usually installed on the low-field side (LFS). Inspired by the result of the high fueling efficiency from the SMBI system at the high-filed side (HFS) on Tore Supra (Tamain et al. 2013), the difference in the transport-related to the magnetic field was further investigated by a 3D edge turbulence fluid code (Tamain et al. 2016), which predicts that an inward flux would be induced owing to the curvature on the high field side, allowing higher fueling efficiency. Thus, HFS fueling can also be practically implemented to improve fueling efficiency.

The deposition of a beam is an important fueling feature to be understood. Experiments on several magnetic fusion devices show that SMBI acts as an effective tool for plasma density and instability control in H-mode plasmas. The deposition of the beam was studied by varying the plasma density increment detected by an interferometer, microwave reflectometry, or reciprocating probe. Measurements from the interferometer on Tore Supra (Bucalossi et al. 2002) and ASDEX-U (Lang et al. 2005) provide the fueling efficiency of the SMBI as well as the influencing time. The reciprocating probe measures the density and temperature profiles, which also helps locate the particle source. The H_α (D_α) array was also used to estimate the injection depth on HL-2A (Yu et al. 2010). From these experimental results, it was observed that the penetration of the SMBI was limited. The physical mechanism has been studied. This suggests that the self-shielding effect may be an important factor. The self-shielding effect might be attributed to the reduction of Franck–Condon neutral penetration due to increased plasma density. In addition to this effect, the plasma profile also has an impact on the injection depth. The depth is related to the operation scenario. A shallower deposition is usually observed in H-mode discharges than in L-mode discharges (Xiao et al. 2012; Yang et al. 2016).

In addition to its good performance in density control, SMBI has also been used as a localized particle source because of its high directionality. Therefore, it has been dedicated to plasma physics studies on various topics. SMBI is usually used to study plasma transport, calculating the transport coefficient using the perturbation transport analysis method as reported in HL-2A (Xiao et al. 2010a) and EAST (Wang et al. 2019). SMBI is particularly effective in studies of the nonlocal heat transport in several devices. The experiments were dedicated to understanding the underlying mechanisms by comparing the proposed theoretical models.

Because SMBI can influence the plasma parameters in the pedestal region of medium-/large-sized devices, it is also used as an external tool to study the H-mode-related physics such as triggering L–H transition and control of the edge localized mode. Experiments on HL-2A have demonstrated that SMBI lowers the power threshold of the L–H transition (Feng et al. 2013; Zhong et al. 2020). In H-mode plasmas, the edge localized modes (ELMs) have been successfully mitigated by SMBI on HL-2A (Xiao et al. 2012), EAST (Zou et al. 2012; Hu et al. 2015), and KSTAR (Xiao et al. 2014b; Kim et al. 2012). Recently, innovations in the SMBI techniques have enabled SMBI to perform well in ELM control with a suitable ratio of the impurity gas to the fueling gas (Zhong et al. 2019).

In addition, the SMBI can flexibly adjust the quantity of the injected particles, which allows particles with a wide range of quantities to be injected into the plasma by changing the pulse duration and source pressure. Thus, the SMBI system meets the requirements of a situation where numerous particles are required, such as disruption mitigation (Huang et al. 2017) and active control of the heat load (Cheng et al. 2013; Gao et al. 2017).

This study presents an overview of the development and the application of the SMBI techniques. A description of the SMBI technique, including the basic features, the testing platform, and the optimization of the beam is presented in Sect. 2. Section 3 introduces plasma fueling using SMBI. The physical study on transport with SMBI is presented in Sect. 4, and the application in the H-mode-related physics is presented in Sect. 5. Finally, the summary and the application prospects are presented in Sect. 6.

2 Description of the SMBI technique

2.1 Basic features

2.1.1 The idealized continuum model

The SMB is formed owing to the adiabatic expansion of the high-pressure gas. Its working background pressure should also be low to ensure few collisions, which is often lower than 10^–4 Pa for the tokamak device. The molecules in the SMB move directionally without colliding with each other. The basic theory can be found in reference (Kantrowitz and Grey 1951), and the experimental proof was subsequently provided (Kistiakowsky and Slichter 1951). Under the assumption of an adiabatic process, no energy transfer occurred outside the system. The conservation of energy, consisting of the enthalpy and kinetic energy of the molecular flow (Zucrow and Hoffman 1976), is:

$$H\left( x \right) + \frac{1}{2}mu(x)^{2} = {\text{constant,}}$$

(1)

where the first term H(x) is the enthalpy on the left side, and the kinetic energy is described by the second expression, $\frac{1}{2}mu(x)^{2}$, in which u(x) represents the average beam velocity. The idealized maximum flow velocity is limited by the total energy of the molecule in the gas source at the temperature of T_0. Assuming all the enthalpy transfer into flow velocity, the idealized maximum velocity is described by the following equation:

$$u_{\max } = \sqrt {\frac{{2H(T_{0} )}}{m}} .$$

(2)

For calorically perfect gases, the heat capacity is not influenced by temperature; thus, the equation of the idealized maximum velocity is rewritten as follows:

$$u_{\max } = \sqrt {\frac{{2C_{{\text{p}}} T_{0} }}{m}} .$$

(3)

For an ideal monatomic gas, $C_{{\text{p}}} = \frac{5}{2}R$ (R is the universal gas constant). From Eq. 3, different gas species can be estimated. The velocity for helium, neon, argon, krypton, and xenon at room temperature are 1770, 786, 559, 386, 308 m/s, respectively. In tokamaks, the velocities of D₂ gas pulses are measured by the D_α emission with fixed flight distance, which is only slightly lower than the velocities estimated above (Yu et al. 2010). This indicates that the assumption of the isentropic expansion is in good agreement with the real situation. The speed of sound (sonic velocity) in an ideal gas is given as follows:

$$a\left( T \right) = \sqrt {\frac{\gamma RT}{m}} ,$$

(4)

where γ is the heat capacity ratio. The SMB velocity limit of an ideal gas can be described as:

$$u_{\max } = \left( {\frac{\gamma }{\gamma - 1}} \right)^{1/2} \left( {\frac{{2kT_{0} }}{M}} \right)^{1/2} ,$$

(5)

where k is the Boltzmann constant, M is the mass of the molecule (atom), and T₀ is the Kelvin temperature of the gas source. The limit velocity of the SMB is equal to [2/($\gamma$− 1)]^1/2 times the sonic velocity. In particular, the velocity of the fueling gas in the tokamak was estimated to be approximately 2.0 km/s for D₂ at 293 K.

2.1.2 Interaction with background gases

The idealized continuum model is valid when the interaction with the background gas is omitted. Considering these interactions, the collisions between the expanding gas particles and the background gas particles would randomize the oriented molecular velocity distribution, leading to a dramatic increase in the entropy.

In a dense background, the gas after expansion soon slowed down in the shock zone, as widely observed by many visualization diagnostics in supersonic gas in the atmosphere. Under a relatively rarefied background condition, the molecules will have sufficient time to experience several other events. The formation of two forms of shock zone, including the barrel-shaped shock and Mach disk, allows the gas to fully expand in a zone of silence, where the velocity becomes larger than the sound speed (M > 1) as shown in Fig. 1. The expansion ended before the Mach disk in the second zone, which manifested as an almost flat shock. The location of the Mach disk was experimentally studied and described using the following empirical equation:

$$X_{{\text{M}}} = 0.67D\sqrt {\frac{{P_{0} }}{{P_{{\text{b}}} }}} ,$$

(6)

where $X_{{\text{M}}}$ is the distance between the outlet of the nozzle and front of the Mach disk. D is the nozzle diameter, and P₀ and P_b are the source and background pressures, respectively. This relation is valid for values of the pressure ratio P₀/P_b in the range of 15–17,000 (Ashkenas and Sherman 1966).

Fig. 1

The features of formation of the SMB

2.2 Cluster

As reported in a previous study (Becker et al. 1956), clusters can be readily formed via supersonic expansions. As shown in Fig. 2, the initial state of the gas was described at point A. The adiabatic expansion of the gas is indicated by the isentropic curve, which crosses the vapor pressure curve at point B. At this point, the state of the gas, together with the kinetic process influenced by the nozzle structure, determines the subsequent evolution trend of the gas. Point C shows the condensation point (Smith et al. 1998), which is further enhanced to obtain a larger cluster under proper conditions. Although no clear prediction of the formation of the clusters has been achieved, the cluster sizes can still be derived by scaling about the nozzle diameter (d), expansion angle (α), and the initial parameters of the gas source, including the pressure (P₀) and the temperature(T₀), which are described as follows:

$$\Gamma^{*} = k\frac{{(d/\tan \alpha )^{0.85} }}{{T^{2.29}_{0} }}P_{0} ,$$

(7)

where k is a constant value related to bond formation. A cluster is regarded to be formed when ${\Gamma }^{*}$> 100–300.

Fig. 2

Schematic plot of cluster formation in free jets. B is the saturation point; C is the onset of the condensation point (Smith et al. 1998)

2.3 SMBI testing platform

To further improve the applications of the SMBI technique in fusion experimental devices, an SMBI testing platform was built for beam analysis and testing of the newly designed SMB injectors. The testing platform has sufficient space and observation windows to be compatible with various testing methods for molecular beams and clusters (Xiao et al. 2021), as shown in Fig. 3.

Fig. 3

Schematic of the SMBI testing platform

The performance of molecular beams and clusters with different parameters and nozzle structures can be tested and optimized using the diagnostic system on the platform, to improve the quality of the molecular and cluster beams, as well as research on injection technology. The entire testing platform includes a vacuum chamber, vacuum pumping system, SMBI system, and an electronic control system. The pumping system is composed of three turbomolecular pumps, which can pump down the chambers at the speed of 1500 L/s. The shape of the vacuum chamber consists of several cylindrical cavities with a total volume of approximately 700 L. Its longitudinal length is 2.2 m which can also be extended as necessary. The maximum and minimum diameters are 500 mm and 250 mm, respectively. The vacuum chamber was carefully processed with stainless steel manufacturing, argon arc welding, polishing surface of special craft processing, and a metal gasket seal between interfaces, which ensures that the vacuum degree can reach 5 × 10^–5 Pa (after baking for degassing). This indicates that the testing platform had a working condition identical to that of the SMBI pipe on tokamaks. The injection system on the test platform has a sliding rail design and a cold trap assembly, which can satisfy a variety of injection conditions. In addition, according to the formula to estimate the quietness zone using Eq. 6, the source pressure and the background pressure are the key parameters that determine the beam characteristics (Yao et al. 2001). Therefore, to ensure the accuracy of the vacuum background pressure measurement of the entire vacuum chamber, two high-precision film gauges with different ranges are installed on the test platform, which can accurately measure the background pressure in the range of 10^–1–10⁴ Pa. The main chamber has adequate windows to satisfy the diagnostic requirements.

2.3.1 Visualization of SMB

IN previous studies, the SMBI technology was investigated by detecting and analyzing the variation of tokamak plasma after SMB, that is, by measuring the plasma parameter changes before and after the injection of the molecular beam. For ages, the performance of a supersonic molecular beam injector was usually evaluated using the fueling effect, which is strongly correlated with the plasma parameters. Thus, it is still difficult to optimize the beam by changing the parameters of the injector, meaning that the uncertainties of the parameters and lack of information on the beam limit the further improvement of the SMBI capability. In recent years, some laboratories have attempted to measure supersonic molecular beams (Liu et al. 2016), however, the accurate measurement of beam parameters is still challenging.

To develop a new SMB injector, the beam was characterized under different conditions such as different Laval nozzle shape sizes, working gas temperatures, and pressures. To directly visualize and measure the transparent gas in the SMBI, a diagnostic system should be established. Commonly, flow visualization technologies can be divided into two categories: One is to analyze the change in the flow field using the change law of light in the variable refractive index field, typical of which are the Schlieren method, shadow method, interferometer. In some cases, quantitative results can be obtained using these methods. The other is to use the tracer to move with the fluid and analyze the flow field structure according to the light scattering or excitation characteristics of the tracer. Common methods include planar laser Mie scattering, particle image velocimetry, planar laser-induced fluorescence, filtering Rayleigh scattering. The working condition of SMBI for the tokamak fueling makes it unlikely to use tracer particles. Considering the simplicity of the optical system, the schlieren method was used to measure the SMBI beam characteristics.

The schlieren system exhibits a refractive index gradient owing to the density gradient in the medium, which is related to the Gladstone–Dale law (Settles et al. 2002). This method has been used in various fields over the last few decades, such as aerodynamics (Barnes and Bellinger 1945), high-speed wind tunnels (Ozawa et al. 2015), and shock waves fields (Estruch et al. 2008). Benefiting from the tremendous boost in the fast imaging industry, schlieren technology has also been well developed to capture fast changes in density, such as supersonic flows (Gonzaleset et al. 2020) and fast processes in air discharges (Zhao et al. 2019). This improvement also makes it feasible to delicately analyze the flow profiles (Medhi et al. 2018).

Numerous designs of schlieren configurations have been created for different application situations. Among these, systems with parallel light rays are preferred for quantitative studies. Based on the parallel light-ray structure, two basic configurations, reflection, and transmission are broadly adopted. On the SMBI testing platform for HL-2A/M, the schlieren system is designed in a ‘Z’ shape, which is suitable for the long focal length of the schlieren mirror chosen for high sensitivity (Xiao et al. 2021). The arrangement of the system is illustrated in Fig. 4. A narrow slit was installed in front of the LED light source to ensure system sensitivity. The light rays passing through the slit are reflected by reflector 1 and collimated by schlieren mirror 1. A plane-turning mirror was placed in the optical path to reflect the parallel light rays into the optical windows of the main vacuum chamber. The SMBI in the main chamber deflects the parallel light rays by disturbing the density, and therefore, the refractive index of the background medium. The deflection angle is relatively small, which ensures that the light rays still travel in a parallel direction. Schlieren mirror 2 refocuses the quasi-parallel light rays with reflector 2 in front of the focal point. A cutter point is placed near the focal point to shed the light rays with certain deflected angles, and the schlieren image with bright and dark areas owing to the shading effect is captured by the high-speed camera.

Fig. 4

Schematic of the schlieren system to measure the SMBI beam (Xiao et al. 2021)

The SMBI beam structure was captured using the schlieren system, as shown in Fig. 5. CO₂ was the working gas with a pressure of 4 × 10⁶ Pa. The background pressure was 10³ Pa. The diameter of the pinhole nozzle was 0.5 mm. From the experimental results in the figure, the molecular beam expands in a paraboloid shape, forming an isentropic region (shock-free silent region), and the disk-shaped Mach shock region in front can be observed. This experimental result was consistent with the theoretically predicted supersonic structure (Ashkenas and Sherman 1966).

Fig. 5

The measured schlieren image of SMB (M is Mach number)

2.3.2 Optimization of the SMBI nozzle

Different applications require different beam characteristics. For fueling, a highly confined gas beam with a high velocity can facilitate deeper deposition of the particles and lower wall retention. Localized injection at the pedestal is preferred for the investigation of H-mode dynamics. In addition, a wide injection with shallow deposition is better for the use of the gas puffing imaging technique. The goal of optimization for fusion devices is mainly focused on improving the beam confinement angle and increasing the beam velocity, which are beneficial to maintaining low gas recycling conditions and deeper penetration. According to Eq. 3, the ideal velocity, derived from idealized adiabatic expansion, is related to the heat capacity, temperature, and mass of the gas. Therefore, the velocity can be increased by an idealized injector structure to realize idealized adiabatic expansion or by simply increasing the temperature of the gas source. The confinement of the beam can be improved by reducing the beam divergence angle and clustering the gas molecular. Clustering the beam requires decreasing the temperature which should be balanced since it also decreases the velocity at the same time. The divergence angle can be changed by changing the injector structure. In this paper, the optimization of the beam means reducing the beam divergence angle. The normal approach to optimization of the injector is firstly to design and investigate the beam characteristics under different injector parameters such as the diameter of the nozzle, the depth of the throat, and the conical angle by fluent simulation (Zhuo and Li 2021; Zhao and Wang 2020). The injectors with smaller divergence angles and large velocities in the simulation would then be manufactured and tested by experiments for further optimization.

An integrated conical nozzle was manufactured based on the simulation result (Xiao et al. 2021). The design diagram and pictures of the normal nozzle and the newly designed integrated Laval nozzle are shown in Fig. 6a, b, respectively. A series of experiments performed on the SMBI testing platform using the Schlieren system show the disparities between the nozzles. In this experiment, CO₂ gas was used, the pressure of the gas source was 40 bar, the background pressure was approximately 100 Pa and the pulse width was 40 ms. Figure 7 compares the beam structure of the schlieren images with the general and optimized nozzles. The divergence angle for the standard nozzle was larger than 90° as shown in Fig. 7a, and Fig. 7b shows that the divergence angle of the optimized case was approximately 30°. A smaller divergence angle results in better orientation properties. Thus, the comparison demonstrates that the directionality is considerably improved after optimization. This result is consistent with the simulation results (Xiao et al. 2021b), in which the integrated nozzle (optimized nozzle) also have smaller beam divergence angle than the general nozzle structure with the same background and gas pressure.

Fig. 6

a Left: cross-section of the nozzle with a pinhole, b Right: cross-section of the integrated conical nozzle

Fig. 7

Comparison of the measured schlieren images for the general nozzle (a) and the optimized nozzle (b) with the same diameter (Xiao et al. 2021)

Several other parameters might also be closely related to the confinement performance of the fueling beam including the length and diameter of the nozzle, according to the theoretical model ⁴¹. This dependence needs to be experimentally validated by further testing experiments, as described in Xiao et al. (2021). The optimization of the beam is only a part that concerns its usage in reactor-like devices. The high magnetic field and the radiation from neutrons also have a strong influence on the present solenoid valve, limiting its usage in reactors. Choices, including enlarging the distance between the valve and device, pneumatic valve with a radiation resistance sealing structure, and other replacing driving forces, should be further explored.

3 Plasma fueling

3.1 Fueling efficiency

The fueling efficiency $\eta_{{{\text{fuel}}}}$ is determined by the ratio of the increase in the plasma content $N_{{{\text{plasma}}}}$ to the injected particles $N_{{{\text{inj}}}}$, as described by the following equation.

$$\eta_{{{\text{fuel}}}} = N_{{{\text{plasma}}}} /N_{{{\text{inj}}}} ,$$

(8)

where $N_{{{\text{plasma}}}}$ denotes the increase in the particles in the entire plasma content, which could be integrated by the variation of density profiles owing to fueling, $N_{{{\text{inj}}}}$ is the injected particles, which can be calculated using the pressure and the pulse of the SMB after calibration. Therefore, the quantitative calculation of fueling efficiency demands the calibration of the particle number. The number of injected particles can be calibrated by detecting the pressure increase in the vacuum chamber or the pressure decrease at the gas source. The number of particles can be calculated by averaging the change in pressure in a certain volume after several pulses of injection.

The SMBI nozzle was carefully calibrated before installation in the experimental devices. For example, a solenoid valve (Parker series 99) was calibrated before its installation on HL-2A. Calibration was achieved by scanning the source pressure and the pulse duration. The quantity of the injected D₂ particles is only related to the source pressure P, with the fixed pulse duration. The number of the particles N with a certain pulse duration of 2 ms, 3 ms, and 5 ms are described using the following formulas:

$$N = \left[ {{0}{\text{.8}} + 7.0P({\text{MPa}})} \right] \times 10^{18}$$

(9)

$$N = \left[ {{1} + 9.8P({\text{MPa}})} \right] \times 10^{18}$$

(10)

$$N = \left[ {{1}{\text{.4}} + 15.1P({\text{MPa}})} \right] \times 10^{18}$$

(11)

The relationship curves are shown in Fig. 8. The above equations can be further concluded, where the particle number N is a function of the pressure and pulse duration, as follows:

$$N = \left[ {{1}{\text{.7}} + 3.{1 }P \cdot t({\text{MPa}} \cdot {\text{ms}})} \right] \times 10^{18} .$$

(12)

Fig. 8

Calibration of the injected particles number of SMBI. The injected particle number versus gas pressure for 2 ms, 3 ms, and 5 ms pulse duration (Chen et al. 2016)

It is considered that the error originates from the counting of the pressure transmitter, which is approximately 5%. In EAST, the number of injected particles was tested on a vacuum test platform using a pressure measurement system (Yuan et al. 2018). A similar relationship is observed.

Several physical processes influence the fueling efficiency because they significantly influence the deposition of the injected particles or edge particle transport. First, the injected particles underwent several processes before reaching the plasma. Only part of the injected particles traveling in the plasma-oriented direction would possibly deposit into the plasma, whereas others would be absorbed by the plasma-facing materials as the wall retention. This process is related to a few aspects, such as the size of the device, the distance between the plasma and the injector, the exhausting ability of pumps, and the divergence of the gas beam. Even plasma-oriented gas molecules can not be injected into the plasma. The molecule was dissociated, ionized in the scrap-off layer (SOL), and exhausted at the divertor. The remainder passing through the SOL could be deposited into the plasma through ionization and collision. The depth of penetration should be highly related to local plasma parameters, such as density and temperature. In general, the plasma density increases, and the density gradient gradually evolves to become steeper after the deposition of the particles in the plasma. The fueling efficiency can be then calculated from the increment of the plasma content over the entire plasma volume. However, it should be noted that the deposited particles may also enhance the edge turbulence transport, especially during a phase of rapid density increase when numerous particles are injected, as will be discussed later in Sect. 4.1.2.

3.1.1 Comparison of fueling efficiency between different fueling techniques

The plasma fueling efficiency is a common concern in magnetic fusion plasma research and is usually defined as the ratio of the plasma density increment to the injected particles as previously mentioned. Owing to the dissociation and ionization processes, the fueling efficiency is expected to also depend on different fueling techniques. Pellet injection is expected to have the highest fueling efficiency with an ablation process to allow a deeper position for dissociation and ionization process, as shown by the experimental results (Takenaga et al. 2009). The fueling efficiency for SMBI is lower than that for pellet injection but higher than GP because the directed motion of the SMB with high velocity enables a higher proportion of the neutral particles to enter the plasma than GP. Sufficient experimental results on HL-2A (Yu et al. 2010), EAST (Zheng et al. 2013), and Tore Supra (Pegourie et al. 2003) suggest that SMBs with higher source pressure and shorter pulse duration, but similar particle numbers, have higher fueling efficiency than GP. In Fig. 9, an SMB pulse and GP pulse are successively injected into the plasma with the same number of injected particles. The maximum increment for the SMBI pulse reaches 0.86 × 10¹⁹ m^-3, whereas the value is 0.34 × 10¹⁹ m^-3 for GP. The density was significantly increased using SMBI. It is expected that the plasma density will firstly increase by ionization from the source and then decreases due to particle transport without sustaining particle source. The D_α monitoring intensity in Fig. 9b indicates that SMBI injected more particles into the plasma. It should be emphasized that huge differences in the fueling rate can lead to different responses of plasma density. A large fueling rate could also strongly modify the local plasma profiles leading to other corresponding behavior such as an opaque area with fueling shielding effect, enhancing turbulent transport, or degradation of the confinement. The fueling shielding effect limits the increase the density, And the enhanced turbulent transport can strong decrease the density to even lower than the density before fueling as observed in Tore supra in Sect. 4.1.2. The profiles of the D_α signal in Figure 10a indicate that the ionization of SMBI injected neutral particles was higher than that of the GP. The time-integrated D_α emission represents the influx of the neutral deuterium particles, which is expected to be proportional to the injected fuel inventory. Without considering the physical processes, the injected fuel inventory of SMBI is larger than GP. Thus, the D_α emission normalized by the injected inventory is used to investigate the effective fuel entering the plasma edge. The fueling efficiency η is calculated from the increase in plasma content over the entire plasma volume in Fig. 10b that. The normalized D_α intensity as well as the fueling efficiency η of SMBI is much higher than those of GP. Thus the normalized D_α intensity can be used to roughly distinguish the fueling efficiency of SMBI and GP since the normalized D_α intensity is much easier to be calculated than fueling efficiency.

Fig. 9

Time evolution of the plasma density (a) and the D_α emission(b) after an SMBI pulse and a GP pulse with similar injected particle numbers (Chen et al. 2016)

Fig. 10

a Profiles of the D_α intensity for SMB and GP. b Normalized D_α intensity versus the fueling efficiency. (Yu et al. 2012)

3.1.2 SMBI fueling efficiency in different confinement regimes

The parameters of the plasma, such as density and temperature, also have a significant impact on the fueling efficiency, as discussed in the last paragraph before Sect. 3.1.1. According to the experimental results reported by EAST (Yuan et al. 2018), SMBI and GP both have higher fueling efficiency in the L-mode than in the H-mode, as shown in Fig. 11. The fueling efficiency appears to properly correlate with the plasma density, which is easily understood as previously discussed. The deposition of the injected neutral particles in a certain plasma area is attributed to collision and ionization processes. A higher plasma density leads to a higher collision rate, and a higher plasma temperature may ease the ionization of neutral particles in outer space, more specifically in the SOL region. The observation that the fueling efficiency decreases with an increase in density is reasonable. Because SMBI has a higher fueling efficiency than GP under different regimes, it could assume the role of GP in fusion devices. Because of its good performance and further optimization in the future (Yao et al. 2004, 2007, 2010; Pegourie et al. 2003; Zheng et al. 2013; Lang et al. 2005; Soukhanovskii et al. 2004; Takenaga et al. 2009; Kim et al. 2013; Murakami et al. 2012; Mizuuchi et al. 2010; Xiao et al. 2014a, 2021; Yao and Baldzuhn 1933; Tanna et al. 2019; Islam et al. 2016), SMBI might play a pivotal role under a variety of conditions for future fusion devices, such as the start-up phase and pellet-free injection of helium discharge.

Fig. 11

Fueling efficiencies of the SMBI system and GP in different plasma discharge regimes (Yuan et al. 2018)

3.1.3 Impact of the far-field SMBI on the fueling efficiency

On Tore Supra, three different fueling techniques were investigated in long-pulse operation (> 200 s), including standard GP, technically complex particle injection, and supersonic pulsed gas injection (SPGI), known as SMBI (Pegourie et al. 2003). The experimental results show that SMBI has a higher fueling efficiency than GP, as observed for all the other devices. The dependence of fueling efficiency on the parameters was investigated, as shown in Fig. 12. The relationship between the fueling efficiency and the distance between the nozzle and the last closed flux surface (LCFS) was obtained, as well as the dependence on the line-averaged electron density and the number of injected particles (Pegourie et al. 2003). As shown in Fig. 12b, c, the number of injected particles rarely influenced the fueling efficiency, whereas the fueling efficiency decreased with increasing electron density, which is consistent with the results of EAST. The influence of the distance D between the nozzle and LCFS on the fueling efficiency is shown in Fig. 12a. As observed, as the distance increases from 5 to 20 cm, the fuel efficiency decreases significantly from approximately 0.8 to 0.4. The results show that the far-field SMBI technology can reduce the fueling efficiency owing to the more significant loss of neutral particles during flight. Fortunately, by optimizing SMB and improving the SMBI system assembly, solutions can be found to enable near-field fueling and further improve the fueling efficiency, as shown by experimental evidence from a divertor SMBI system with its injector near the divertor target (Xiao et al. 2021). In this report, the valve outside the vacuum chamber and the injector near the divertor target inside vacuum chamber are totally separated. The reliability of the valve can be guaranteed when it is far away from the magnetic coil outside the vacuum chamber, and the performance of the beam can be ensured by the laval like injector near the plasma.

Fig. 12

Relationship between the fueling efficiency (${\varepsilon }_{\mathrm{fuel}}$ is the ratio of the increase of the plasma content divided by the number of injected particles just at the end of the gas injection.) and a the distance Δ between the nozzle and the LCFS, b the line-averaged plasma density n_l, and (c) the injected particle number (Pegourie et al. 2003)

3.2 Density feedback control

3.2.1 Time response of SMBI on density feedback control

Plasma density control is challenging for fusion devices with high-performance and long-pulse operation. The GP and SMBI systems are both equipped on the HL-2A. The nozzle of the LFS mid-plane SMBI system is approximately 1.5 m away from the plasma edge, and the GP piezoelectric valve is similar. Figure 13 compares the plasma responses after feeding with GP and SMBI. The delay time between the increase in the D_α signal and GP triggering was approximately 8 ms, whereas the delay time of SMBI was only approximately 0.7 ms. The rapid response capability of SMBI indicates that SMBI is an effective approach for density feedback control.

Fig. 13

The discrepancy of the time response between gas puffing (left column) and SMBI (right column) (Chen et al. 2016)

Figure 14 shows a typical plasma density feedback control experiment using SMBI. The preprogrammed plasma density ccn_e is represented by the red line, and the measured density n_e is represented by the blue line. The SMBI gas pressure in this discharge was set to be 1.0 MPa depending on the density supply requirement. The density feedback control system controlled the SMBI injector by generating successive SMB pulses of different durations. The plasma density was in good agreement with the default value of the plasma density. In general, rapid feedback control means fewer particles need to be injected. Thus, the wall retention was much lower than gas expansion.

Fig. 14

Comparison of the experimental result of density feedback control using the SMBI system to the preprogrammed value of density ccn_e (Chen et al. 2016)

3.2.2 Density feedback control for long-pulse devices

The plasma density feedback control of the steady-state operation is crucial for a long-pulse superconducting tokamak. In EAST, the density feedback control system also integrates the SMBI system as the actuator, and its control method is similar to that used for GP (pulse-width modulation) (Zheng et al. 2013). Figure 15 shows a typical discharge of the density feedback controlled by the SMBI on EAST. During this discharge, pre-filling is achieved by a long pulse of GP before 1.2 s, and then the density is controlled by the SMBI system. The density increased linearly, and the accuracy of the density feedback control was high, with an error of only 3% relative to the reference density. In addition, experimental results show that the 0.7 Greenwald density limit has been achieved by SMBI density feedback control.

Fig. 15

SMBI density feedback experiment by SMBI on EAST (Zheng et al. 2013)

The effects of GP and SMBI on wall retention were studied. Figure 16 shows two similar discharges with feedback controlled by GP and SMBI. The default density values for the two discharges were similar. As observed, the amount of the fueling gas injected by SMBI was reduced by 30%, and the exhaust inventory was only reduced by about 14%, compared to the case of GP. Therefore, the wall retention rate, which is the difference valve between injected particles and the exhausted particles in the SMBI cases, was reduced by 43% owing to the high fueling efficiency of the SMBI. It also shows that SMBI has the advantage of lower wall retention for long-pulse plasma discharge than GP. Because wall retention has a significant impact on plasma performance, SMBI with low gas recycling is desirable. The optimization of the beam divergence angle is expected to significantly reduce the gas diffusion, and further improves the performance of the SMBI on maintaining low gas recycling condition.

Fig. 16

Density feedback discharges with SMBI and GP (Zheng et al. 2013). a line-averaged electron density, b the amount of the fueling gas, c exhaust inventory rate

In a long-pulse discharge experiment on Tore Supra (Tsitrone 2003), the SMBI system was also successfully implemented to control the plasma density with a feedback control loop. Compared to GP, the amount of gas required for SMBI fueling discharge is reduced by approximately 20–30%. In addition, the wall inventory was significantly reduced by nearly 50%.

4 Studies on plasma transport

4.1 Particle transport

4.1.1 Perturbation transport analysis by SMBI

Plasma confinement is highly related to particle transport, particularly in the radial direction. Periodic GP is typically used to study particle transport physics using the perturbation transport pulse propagation method. However, non-directional GP often leads to high recycling, as much puffed gas is retained on the wall. Therefore, it may be challenging to achieve a constant density region to study particle transport using the perturbation method. A new gas-injection approach with better directionality and lower wall retention is desirable. Experiments on HL-1 M have already demonstrated that SMBI is suitable as a particle perturbation source because it can deposit in a deeper and more localized region in the plasma than GP.

The key point for the particle transport study using the perturbation method is to maintain the baseline of the averaged density almost unchanged. To ensure this, wall conditioning was used to reduce wall retention, and GP was not supposed to be injected. In addition, the SMBI pulse parameters should be carefully optimized to obtain a flat density period.

Particle transport was investigated by modulation of SMBI in HL-2A. Figure 17a shows the density evolution at different radii as detected by microwave reflectometry. The measured relative density perturbation is approximately 15% of the background density. The ratio of the electron temperature to the local electron temperature was approximately 12%, as indicated by the electron cyclotron emission (ECE) signal in Fig. 17b. The variation in the average density and temperature did not strongly influence the particle transport studies.

Fig. 17

Time evolution of a plasma density at different plasma radius with modulated SMBI on HL-2A b electron temperature to the SMBI (Xiao et al. 2010a)

Several parameters should be obtained before calculating the particle transport coefficients, such as the deposition of the particle source and the boundary of different domains with different features. The location of the particle source can be roughly resolved by varying the D_α or density signal. The particle source can also be obtained by Fourier transform (FT) of the modulated density owing to SMBI, where the location of the source corresponds to the location of the minimum of the phase. The results of the fast FT also provided information on the phase and amplitude. Boundaries can be roughly defined according to the change in phase and amplitude. The particle diffusivity D and convective velocity V in different regions can be calculated by fitting the experimental data with an analytical model (Xiao et al. 2010a). Figure 18 shows the experimental and fitting results of the simulation. The results indicate that the phase is prone to diffusivity but is much less sensitive to the convective velocity. Whereas the amplitude is the reverse, which significantly depends on the convective velocity.

Fig. 18

Sensitivity of D and V on the amplitude and the phase of the first harmonic of the FT of the modulated density (a) with fixed V, and (b) with fixed D. (The boundary between domains I and II is defined by the position of the source at r = 27.5 cm, and the boundary between domains II and III corresponds to an obvious change in the profile of the amplitude at r = 30 cm.) (Xiao et al. 2010a)

The coefficients obtained using perturbative analysis are in a transient state. Distinct differences may exist in the coefficients obtained using the power balance for the quasi-steady state. In addition, the transient flux gradient can be easily calculated using the transport coefficients (Zhang et al. 2020).

Owing to the relatively deep deposition at the plasma edge, the modulated SMBI can also be used in particle transport analysis in H-mode discharge. ELMs can strongly influence the pedestal density, which has a significant impact on the calculation of particle source deposition. In addition, the SMBI modulation frequency should be distinguished from the ELM frequency to reduce the impact of ELM crashes on the modulated particle source.

Figure 19 shows the particle transport study with SMBI modulation in H-mode discharge (red line), as well as L-mode discharge (blue line) and the discharge with only Ohmic heating (black line) (Wang et al. 2019). As observed from the density signal, the density is modulated by SMBI pulses. The calculated transport coefficients for the different cases are shown in Fig. 20. The diffusion coefficients D and convective velocity V were obtained by the density modulation with SMBI pulses, as shown in Fig. 20a, b. The features of different regimes are explained using the results, such as the strong inward convection owing to the edge transport barrier in the H-mode. It is worth emphasizing that the diffusion coefficient is higher when considering the constitution of ELMs, which indicates enhanced particle transport by ELMs in the periphery.

Fig. 19

SMBI modulation experiment in the Ohmic discharge (black), L-mode discharge(blue), and the H-mode discharge (red) (Wang et al. 2019)

Fig. 20

a Diffusion coefficients (D_m) and b convection velocity (V_m) from modulation experiment in different discharge regimes (Wang et al. 2019)

4.1.2 Particle transport behavior by SMBI fueling

The particle transport is commonly observed to be enhanced during fueling experiments by SMBI as described in Sect. 3.1. In Tore Supra, particle transport is investigated by changing the plasma current (Tamain et al. 2013). The process of the fueling effect was separated into four phases, as shown in Fig. 21. The first phase shows a rapid increase in density, followed by a rapid decrease in the density in phase two. The excessive declined density in phase two recovers and saturates in phase three. In phase four, the density slowly decays back to the status as in phase one. Experimental observations show that phases two and three are strongly affected by plasma current. The duration of these two phases becomes shorter with the increase of the plasma current. The fueling efficiency referred to the density at the end of the third phase is also lower, when the plasma current is lower. The measurements of the edge turbulence by the high-speed visible camera and the Langmuir probe array suggest that dramatic change of the edge fluctuation by SMBI occurs with a slower variation of larger coherent structures. These large coherent structures of turbulence lead to a transient period of confinement degradation with higher outwards particle transport, thus lower fueling efficiency. However, it should be noted that the possible dependence of plasma density is not excluded from these results. The effect of Greenwald fraction on the fueling process can be further exploited.

Fig. 21

Time evolution of the density at different plasma currents in Tore Supra (Tamain et al. 2013)

More attempts have been made to shed light on further interactions between SMBI and edge plasma. Simulations of the transport dynamics with SMBI were conducted by BOUT +  + , considering the particle, heat, and momentum transport equations for different particles including ions, atoms, and electrons (Wang et al. 2014). The simulation indicates that the peak of plasma density and fronts of the neutral particle flow propagate inward continuously and then move outwards during SMBI in the radial direction. Along with the local peak of the plasma density, the electron temperature is locally reduced in the poloidal direction. Furthermore, based on this module, the self-shielding effect of the first ionized SMB and fueling penetration depth of SMBI affected by plasma density and temperature profiles were studied (Shi et al. 2017a; Wu et al. 2017). Subsequently, a new seven-field two-fluid model in the BOUT +  + framework is also developed to simulate SMBI fueling into H-mode deuterium plasma on the EAST (Qian et al. 2020). The simulation results indicate that SMBI has a self-shielding effect on the inward transport of molecules, which is in good agreement with the experimental results (Yu et al. 2010).

Experiments and simulations also show that the particle transport behavior by SMBI is different between LFS and HFS. The fluctuations detected by the probes on the LFS midplane and the reciprocating probes at the top of the Tore supra support the enhanced turbulence owing to SMBI, as described previously. As shown in Fig. 22, the ion saturated current of the LFS is higher than HFS indicating larger outward particle transport in the LFS. Although the evolution of local fluctuation is influenced by the movement of the probes. It can still be noticed that the turbulent fluctuation in both LFS and HFS after fueling is enhanced since the increase of ion saturated current on the way back of the probe is much slower than the increase on the way in. The plasma dynamics after SMBI at LFS and HFS are analyzed using the TOKAM3X 3D edge fluid turbulence code (Tamain et al. 2016). Simulations suggest that the plasma behaves differently in LFS and HFS injection due to the different curvature signs. The favorable curvature at the HFS triggered the inward particle flux, and a large proportion of the particles flowed from the deposition source to the plasma core. Thus, the fueling efficiency in HFS ranges from 50 to 70%, which is much higher than the efficiency in LFS ranging from 30 to 45% (see Fig. 23).

Fig. 22

Time evolution of the ion saturation current during an SMBI pulse at 1.2 MA plasma current (Tamain et al. 2013)

Fig. 23

The pITB is enhanced by a pulse of SMB in HL-2A. a Plasma current Ip, the central averaged electron density. b Mirnov coil signal and H α signal. c Time–space evolution of the density gradient (Xiao et al. 2010b)

4.1.3 ITB control by SMBI

The internal transport barrier (ITB) has been widely investigated because it is accompanied by a sharp pressure gradient in the core and enhances the bootstrap current with the growth of the non-inductive current ratio. ITB is considered to sustain its large pressure gradient layer by reducing turbulence via radial electric-field shear or magnetic-field shear (Koide et al. 1994; Gormezano et al. 1998). Particle and torque inputs might benefit the formation of ITB. Thus, the active and effective control method of ITB from the accessible edge region is desirable. Experimental results indicate that the application of SMBI eases the achievement of an ITB.

A spontaneous particle internal transport barrier (pITB) was first demonstrated in Ohmic plasmas on HL-2A (Xiao et al. 2010b). In this experiment, no external momentum input was applied. Only GP was used for pre-fueling. However, evidence shows that the pITB formation does not correlate to gas puffing but could be enhanced by SMBI injected at 510 ms as shown in.

The observed results suggest that the pITB is robust and sustained after a pulse of SMBI. Transport analysis shows that the convective velocity changes sign after the injection of SMBI. Consequently, the inward particle flux after SMBI contributed to the enhancement of the pITB. The directional change in the convective velocity can be interpreted by the change in turbulence characteristics. Before SMBI, the trapped electron mode(TEM) turbulence is dominant. The injection of the SMBI increases the local density which leads to an accession of the density threshold, the transition from the TEM with outwards convective velocity to ion temperature gradient mode (ITG) with the occurrence of inward convective velocity.

Particle control experiments with ITB towards the burning plasma control were performed in JT-60U. To improve the controllability, SMBI was demonstrated to have a quicker response than GP. It was installed to enhance the particle control capability (Takenaga et al. 2009). GP, SMBI, and PI were all used to investigate the impact on plasma confinement as shown in Fig. 24 (Takenaga et al. 2009). The plasma confinement is represented by the H-factor using the ITER89P L-mode scaling. Its dependence on the normalized density is shown in Fig. 24a. It could be observed that the accessible density range is extended to n_e/n_GW = 0.7. Confinement degradation by GP is attributed to its significant impact on the T_i ITB. However, the confinement is not affected using the pellet injection which still maintains the T_i ITB. A moderate degradation of the confinement is also found using SMBI. The difference between these fueling techniques could be possibly explained by the different depositions and their effect on the parameter profiles. The shallower deposition of the GP leads to a reduction in the electron temperature and the pressure gradient at the plasma edge as shown in Fig. 24b. This also decreases the temperature; however, the temperature soon recovers, leading to an increase in the pressure gradient. The SMBI is an intermediate, that can be deposited in different areas by adjusting the injection parameters. The pellet is estimated to penetrate the pedestal top at the normalized radius to be 0.77–0.84. The change in plasma confinement is closely related to strong core–edge coupling. In the GP cases and some SMBI cases, shallow deposition results in a decrease of the edge pressure gradient with higher particle transport possibly weakening the core confinement. However, in other SMBI cases and pellet injection cases, the deep penetration of the particles increases the pressure gradient, which improves the core confinement. After optimizing SMBI parameters such as the directionality and speed, a deeper deposition could be achieved in the experimental devices for the control of plasma particle transport.

Fig. 24

a H_98PL versus Greenwald fraction b nT-diagram at the pedestal (Takenaga et al. 2009)

Further investigation suggests that the modification of the parameter profiles by SMBI is not only related to deposition, as evidenced in JT-60U. Recently, SMBI experiments with ITB sustained by weak magnetic shear have been conducted to investigate the ITB dynamics in response to the cold pulse induced by SMBI(Kin et al. 2021). The SMBI pulses were indicated by a prompt increase of the line-averaged density. The spatial–temporal evolution of the inverse scale lengths of T_i and T_e separately are shown in Fig. 25b, c. Repetitive cold pulses were induced with similar SMBI parameters, which never led to the same response on the temperature gradient. Instead, both the ion and electron gradients increase and decrease alternately with the SMBI, and the ITG responds more strongly to SMBI. The increase represents the steady ITB phase, and the decrease corresponds to the ITB recovery phase. The difference in the transport features during different ITB development phases could cause alternative changes in the temperature gradients. The experimental result indicates that it is possible to regulate the plasma temperature in the core region by particle injection at the edge, particularly where direct access to the plasma core region is severely limited.

Fig. 25

a Line-integrated density; spatial–temporal evolution of the inverse scale length of b ion temperature and c electron temperature (Kin et al. 2021)

4.2 Heat transport

Nonlocal heat transport (NLT) has attracted broad interest from researchers since it was first observed several decades ago and is also a challenging and unsolved issue in plasma physics (Gentle et al. 1997). NLT manifests as the rapid growth of the electron temperature in the core plasma activated by the edge cooling, especially with low electron density.The observation of the NLT phenomenon indicates that it is not sufficient to explain the observations by the local transport model. Significant efforts have been devoted to understanding the underlying physics mechanisms using different approaches, such as impurity seeding by the laser blow-off technique, pellet injection, and other particle injection techniques. A nonlocal effect was achieved using SMBI in the HL-2A tokamak as shown in Fig. 26. It can be observed that the outside region of plasma is cooled by an SMB pulse, while the core electron temperature T_e increases about 18% for a duration of approximately 40 ms.

Fig. 26

Time evolution of electron temperature T_e at the different radius (Sun et al. 2009)

Further analysis of heat transport with SMBI modulation was performed (Sun et al. 2008, 2009). Sequential SMBI on the HL-2A tokamak was used to sustain the growth of the core temperature in response to edge perturbations (Sun et al. 2011). It has been observed that the central turbulence is suppressed during the nonlocality from the measurement of O-mode reflectometers as shown in Fig. 27, suggesting that the interpretation of this phenomenon is reasonable owing to the formation of an "ITB-like" structure.

Fig. 27

SMBI-induces nonlocal effect on electron temperature. From top to bottom: the core and edge electron temperature, line-averaged density and cut-off surface of the reflectometer, and spectrogram of density fluctuation (Sun et al. 2011)

Further studies have shown that the nonlocal effects caused by SMBI are associated with beam deposition (Sun et al. 2010). For low densities, SMBI performed better in HL-2A to sustain the increase in core temperature than the nonlocal phenomena caused by GP. The results show that when the density is above a certain limit, the nonlocal effect usually disappears. In Alcator C-Mod, it is observed that the NLT is related to the confinement regime and the direction of intrinsic rotation. Rises in the temperature are observed only in the linear Ohmic confinement (LOC) regime where the plasma density and the collisionality are lower. At the higher density, the NLT disappears and a temperature drop is observed (Rice et al. 2013). More experimental evidence from KSTAR suggests that the plasma density directly limits nonlocal transport. An intuitive and clear structure of the temperature fluctuation can be observed using an ECE imaging system in a two-dimensional space (Shi et al. 2017b).

Several theoretical models have been proposed to explore the physics of nonlocal transport. The self-organized criticality (SOC) paradigm proposes a relationship between large-scale transport events and individual turbulent eddies through avalanches by the “sandpile” model. The observation of HL-2A revealed that the SOC dynamics play an important role (Pan et al. 2015). This result is supported by the enhanced avalanche behavior in the early nonlocal stage as shown in Fig. 28. It can be observed that the Hurst parameters, self-similarity in turbulence, and large-scale radial correlation are all enhanced with the growing inward propagation of turbulent activities.

Fig. 28

Comparison of SOC features without NLT (black) and with NLT (red). a Hurst exponents at r/a = 0.77; b radial dependence of Hurst parameters; c PDF of large-scale transport events versus the effective velocity of avalanche for inward (V_r < 0) and outward (V_r > 0) propagation (Pan et al. 2015)

In the transient NLT event, the interaction between the nonlocal transport and the neoclassical tearing mode (NTM) was observed(Ji et al. 2016). A typical experiment for the interaction study is shown in Fig. 29. From the spectrogram in Fig. 29c, it can be observed that the NTM is excited nonlocally. The excitation is mainly caused by the transient increase of local pressure gradient during nonlocal transport caused by SMBI on q = 3/2 rational surface. In return, the NTM can regulate the nonlocal transport by truncating avalanches through sheared toroidal flows near the magnetic island.

Fig. 29

Typical discharge waveforms with an onset of a 3/2 NTM during the nonlocal transport induced by SMBI (vertical yellow bar) in ECRH-heated plasma at HL-2A (Ji et al. 2016)

Turbulence spreading is proposed as a candidate mechanism for the interpretation of nonlocal transport. Recently, evidence has been found using SMBI, which supports the turbulence spreading mechanism in the nonlocal transport experiments in EAST Ohmic plasma (Liu et al. 2019). After 10 ms of the SMBI, the electron temperature started decreasing, and the core density started growing. Subsequently, the electron temperature reached the maximum point of about 30 ms, as shown in Fig. 30. It is also observed in Fig. 30 that the core electron density fluctuation started increasing within 5 ms after SMBI, before the increase in the core electron temperature. Its amplitude reaches a maximum value around 15 ms after SMBI and then recovers back to the same level of amplitude before SMBI within roughly 50 ms. This result indicates that there is another aspect influencing the turbulence intensity in addition to the local gradients, however, the turbulence spreading model might better meet the observation. In practice, the increase of the electron density fluctuation level could be regarded as a sign of turbulence enhancement after SMBI.

Fig. 30

Evolution of core electron temperature, electron density, and electron density fluctuation at a similar location after SMBI. (Liu et al. 2019)

5 H-mode related dynamics with SMBI

5.1 Stimulated L–H transition

Investigation of the L–H transition is crucial for the operation of the H-mode for the fusion devices, such as ITER and DEMO, whose desirable power threshold of the H-mode should be low to obtain the expected energy gain factor Q. Sufficient evidence shows that the power threshold for the L–H transition is sensitive to many parameters and physical processes which are included or not included in the scaling law, such as the X-point height(Gohil et al. 2011; Meyer et al. 2011), triangularity δ (Andrew et al. 2004, 2008), wall conditioning (Ryter et al. 2013; Maggi et al. 2014), and isotopic effect (Ryter et al. 2013; Righi et al. 1999; McDonald et al. 2004). Although plenty of parameters can affect the L–H transition process, the nature of L–H transition is well accepted to be the formation of the edge transport barrier due to turbulence suppression (Burrell 1997). The L–H transitions triggered by the magnetohydrodynamics (MHD), such as sawtooth (Kobayashi et al. 2015; Shao et al. 2020) and fishbone (Liu et al. 2014) have been observed. The heat flux released during MHD activities from the core to the edge can enhance the poloidal shear rate by increasing the plasma pressure gradient or the turbulence Reynolds stress. Similarly, the L–H transition with pellet injection also indicates that the variation in radial electric-field shear owing to density fluctuations and ITG should be responsible for the turbulence suppression (Belokurov et al. 2018). In addition, fueling effect on transition power threshold P_th has been observed with pellet injection (Gohil et al. 2001; Yao et al. 2017). Recently, results from L–H transition experiments on HL-2A also show that SMBI (Feng et al. 2013; Zhong et al. 2020; Duan et al. 2010) has an impact on the L–H transition power threshold. Figure 31 shows the correlations between the statistical power loss P_Loss of the L–H transition with the line-averaged density in the HL-2A experiments with or without SMBI. P_Loss is defined as:

$$P_{{{\text{Loss}}}} = P_{{{\text{OH}}}} + P_{{{\text{AUX}}}} - {\text{d}}W_{{\text{E}}} /{\text{d}}t,$$

(13)

where P_OH stands for the ohmic heating power, P_AUX represents the auxiliary heating power, only NBI in these discharges, and dW_E/dt is the time derivation of the stored plasma energy. The experimental result has shown that the relation between the power threshold and density is non-monotonic (Snipes et al. 1996), which is explained by the role of the ion heat flux (Ryter et al. 2014). From the experimental observations on HL-2A, a similar result has been found. It is observed that the power threshold for the low-density region (< 2.0 × 10¹⁹ m⁻³) is significantly decreased with the plasma density (red triangles in Fig. 31), while in the high-density branch, the power threshold increases with the increase of density. The experimental analysis also suggests that the minimum threshold is determined by the ratio τ_E /τ_ei (energy confinement time/volume-averaged electron–ion energy exchange time)(Sauter et al. 2012).

Fig. 31

Plots of the power loss versus line-averaged electron density in H-mode with or without SMBI (Zhong et al. 2020)

To further explore the correlation between SMBI and the L–H transition process, the injection time of SMB was modified for plasma with identical discharge parameters on HL-2A. Figure 32 shows two typical discharges with P_NBI = 1 MW, the heating power in these two cases is close to the marginal H-mode power threshold of the HL-2A (Zhong et al. 2020). From Fig. 32(e), it can be observed that the SMB is injected at different times and triggers the L–H transition after 10 ms, meaning the transition for shot 27,887 is postponed because of the postponement of the SMB injection time. Further analysis suggested that the L–H transition was not caused by the fueling effect (Zhong et al. 2020). The results in Fig. 32 indicate that the L–H transition occurs with different densities in these two cases, which further indicates that the stimulated transition does not result from the increased density modified by SMBI.

Fig. 32

Wavefront of two similar H-mode discharges stimulated by SMBI at different times (Zhong et al. 2020)

In these cases, the transition is not caused by the fueling effect of the SMBI. Another plausible mechanism related to the flows and turbulence may be at play, as observed in experiments. The SMBI can influence the plasma flow, turbulence, and the underlying transport, and therefore, influence the confinement. Previous theoretical studies have suggested that turbulence can be excited in the transport barrier region (Han et al. 2017).

As observed in the experiments, the turbulence evolution measured by Doppler back-scattering (DBS) reflectometry suggests that SMBI destabilizes the turbulence. The coherent mode during the SMBI is identified as the geodesic acoustic mode (GAM) by correlation Doppler reflectometry (Zhong et al. 2015). During the L–H transition with SMBI, the GAM regulates the turbulence to form a steeper edge density profile. The SMBI increases the plasma by increasing the edge plasma density and decreasing the edge plasma temperature. A higher collisionality leads to a strong coupling effect between electrons and ions, which results in a larger ion diamagnetic term of E_r according to the radial force balance equation. The turbulence suppressed by a larger E_r lowers the requirement in the L–H transition process.

In conclusion, it has been demonstrated that SMBI can trigger the L–H transition by lowering the H-mode power threshold. Similar to other triggering approaches, the turbulence was also suppressed during the L–H transition with SMBI. Differences were observed between turbulence suppression during L–H transition by SMBI and by other approaches such as pellet or MHD events. The pellet can significantly change the density gradient and further enhance the radial electric shear as reported in experimental result (Belokurov et al. 2018), in which the mean flow could be strengthened and suppress the radial transport events. The L–H transition triggered by MHD events may also increase the pressure gradient by releasing heat flux from the core to the edge area, and the Reynolds stress-induced zonal flow might also be enhanced, leading to turbulence suppression (Liu et al. 2014; Belokurov et al. 2018; Gohil et al. 2001). However, it has been observed that the enhancement of nonlinear interactions between shear flows and turbulence, such as the interactions between the SMBI-stimulated GAM and the ambient turbulence, should play a key role in the L–H transition with SMBI. During the experiments of the L–H transition triggered by SMBI, it has been found that the nonlinear interactions between mesoscale flow and turbulence are enhanced. The enhanced nonlinear regulation dynamics can quench the ambient plasma turbulence and maintain the turbulence collapse, resulting in an L–H transition. For these characteristics, SMBI can be a potential approach to control the process of L–H transition around the marginal heating power of the H-mode in magnetic fusion devices.

5.2 ELM mitigation

Since the first operation of the H-mode discharge, it has been envisaged to be a standard operation scenario for future fusion reactors. The H-mode is also well known and characterized by the ELM, which periodically ejects a large number of particles and heat onto the plasma-facing components (PFCs), especially the divertor. ELMs are divided into many types according to the characteristics such as frequency or amplitude. Type-I ELM may lead to a large collapse at the pedestal and cause significant damage to PFCs, which must be mitigated by external methods in the fusion devices. Research on several small ELM regimes, such as type-II ELM and grassy ELM, is also underway. This aims at realizing high plasma performance and small erosion of PFCs (Viezzer 2018). It is essential to determine the mechanisms of different ELM types. Normally three types of ideal MHD instabilities driven by the steep current and pressure gradients at the edge transport barrier (ETB) are supposed to explain the formation of various ELMs (Liang 2015): peeling mode, ballooning mode, and the peeling–ballooning mode. Type-III ELMs locate at the peeling boundary as commonly observed in HL-2A, type-I ELMs locate at the ‘nose’ part of the peeling ballooning (P-B) boundary, while type-II ELMs always occur with high collisionality and are located at the ballooning boundary (Snyder et al. 2002).

Several techniques have been successfully adopted to control large ELMs, such as pellet injection (Lang et al. 2003), resonant magnetic perturbation (RMP) (Evans et al. 2006), and radio-frequency wave injection (Liang et al. 2013). The commonly observed ELM mitigation or suppression could be interpreted by the P-B model. On the one hand, the current and pressure profiles can be modified to be located below or beyond the peeling ballooning boundary by heating, current drive, and particle injection techniques (Lang et al. 2013). On the other hand, the peeling ballooning boundary may be enlarged by modifying the plasma shape or E_r shear (Snyder et al. 2007). In addition, other pedestal instabilities such as turbulence or quasi-coherent mode can also mitigate or suppress ELMs by nonlinear interaction which dissipates the pedestal energy to restrain the appearance of large ELMs (Xiao et al. 2019).

SMBI has also been demonstrated to be an effective and flexible approach to control ELMs on HL-2A (Xiao et al. 2012, 2014b; Yang et al. 2016), KSTAR (Xiao et al. 2014b; Kim et al. 2012)^.and EAST (Hu et al. 2015) (Wan et al. 2015), in which the SMBI-induced minor instabilities such as turbulence and high-frequency bursts can prevent large ELMs. Intensive studies to understand the physical mechanism of ELM control. In this section, the result of the ELM control with different SMBI techniques such as deuterium SMBI, impurity SMBI, and mixture SMBI is introduced.

5.2.1 Observations

D₂ fueling of SMBI was first observed to be an effective approach for ELM control during type-III ELMy H-mode discharge on the HL-2A tokamak (Xiao et al. 2012). It has been successfully applied in EAST (Hu et al. 2015) and KSTAR (Xiao et al. 2014b). Generally, thresholds of the SMB parameters had an impact on ELMs. For the ELM mitigation experiment with the fueling SMB on HL-2A, the gas source pressure is often chosen to be larger than 1 MPa, and the pulse duration is longer than 2 ms. Figure 33 shows the ELM mitigation for a typical H-mode discharge with fueling SMBI on the HL-2A tokamak. It was observed that the ELMs were mitigated after SMBI injection with an increase of the line-averaged density. The time of the ELM mitigation phase (or the SMBI influence time) τ_I was about 20 ms.

Fig. 33

ELM mitigation experiment with SMBI on HL-2A a the SMBI valve signal, b the Dα signal, c the time interval between ELMs, d the ELM amplitude, e the total energy confinement time, and f is the line-averaged electron density (Xiao et al. 2012)

Figure 34 is the ELM frequencies f_ELM with SMBI and without the SMBI. It can be observed that the frequency ${f}_{\mathrm{ELM}}^{\mathrm{SMBI}}$ is larger than ${f}_{\mathrm{ELM}}^{0}$, which denotes the mitigation effect of the SMBI. Figure 33e shows an almost unchanged total energy confinement time, which indicates no degradation of the plasma confinement.

Fig. 34

ELM frequency versus the pma current ELM frequency f_ELM with and without SMBI. The increase factor in frequency of ${f}_{\mathrm{ELM}}^{\mathrm{SMBI}}$/${f}_{\mathrm{ELM}}^{0}$ is about 2–3.5 (Xiao et al. 2012)

Because large ELMs are mitigated, the associated heat load on the divertor induced by ELMs decreases, as shown by the heat flux distribution detected by the infrared camera on HL-2A in Fig. 35. The heat flux detected by Langmuir probes from KSTAR (Xiao et al. 2014b) suggests that the heat flux at the inner divertor is much lower than that at the outer divertor and does not change significantly during the ELM mitigation phase.

Fig. 35

The divertor heat flux Q_div evolution with ELM mitigation. a The D_α signal b divertor heat flux distribution (Xiao et al. 2014b)

5.2.2 Parameter dependence

To successfully perform the ELM mitigation experiments with fueling SMBI, it should be noted that several parameters should play important roles. The gas pressure and pulse duration should be selected according to the discharge parameters in different devices as reported in the above section. In general, the effect of SMBI on the ELMs is closely related to the SMBI pulse. In addition, the thresholds for the SMBI parameters differ for different devices based on the results of HL-2A and KSTAR (Xiao et al. 2014b). Thus, optimization of parameters of SMBI has a priority in conducting the ELM mitigation experiments in a certain device. In addition, the plasma density, energy confinement time, and deposition of the SMB can influence the behavior of the ELMs.

The dependence on the plasma density and energy confinement time was compared between HL-2A and KSTAR (Xiao et al. 2014b). From the experimental results for HL-2A, it was observed that the SMBI influence time τ_I is about 15–25 ms. The influence time was approximately 250–400 ms in KSTAR. Compared with the energy confinement time, there was a difference. τ_I in HL-2A is close to τ_E in HL-2A (τ_E ~ 20–30 ms); however, τ_I in KSTAR is much larger than its τ_E (τ_E ~ 120–170 ms) (Yoon et al. 2011), as shown in Fig. 34. The difference in density is significant for these two machines. In HL-2A, the line-averaged electron density was about 2 × 10¹⁹ m⁻³, whereas the line-averaged density is about 3.5 × 10¹⁹ m⁻³ in KSTAR. According to Figs. 36a, 34b, the influence time τ_I is close to the energy confinement time τ_E in the lower density region (τ_I/τ_E ∼1), whereas the influence time is longer than the energy confinement time τ_E in the high-density region (τ_I/τ_E ~ 2–3). The difference in different density regions might be related to particle transport, as indicated by the theoretical studies (Takenaga et al. 1995). The results show that the particle confinement time is comparable with energy confinement time in the low-density regime with a ratio of approximately 1, which is assumed in the linear confinement region. However, for the high line-averaged density regime, this ratio is larger than 2, which agrees with the experimental observations on the influence. The result suggests that the SMBI influence time τ_I on ELMs is determined by the particle transport events.

Fig. 36

τ_E, τ_I and τ_I/τ_E in HL-2A and KSTAR. a Experimental results of τ_I and τ_E and b is the ratio of τ_I/τ_E in HL-2A and KSTAR

Intuitively, the ELM mitigation effect should also depend on the deposition of the particles injected by the SMBI system. The deposition location of the particle source by SMBI was roughly estimated by the maximum value of the time derivation of the density (dn_e(r)/dt) at different radii. The x-axis in Fig. 37 is the radius position normalized by the width of the pedestal, which is about 2-3 cm for a typical H-mode discharge on HL-2A. Therefore, the pedestal bottom is defined as 0 and the pedestal top was 1. The red upper triangle represents the ratio of the ELM frequency between the mitigated ELMs and natural ELMs, which is described as F_{ELM mitigated}/F_{ELM natural}, the ratio of energy lost per ELM between the mitigated ELMs and natural ELMs is marked by the blue lower triangle described as E_{ELM mitigated}/E_{ELM natural}, and the ratio of the ELM-induced heat flux between the mitigated ELMs and natural ELMs is marked by the green circles described as H_{ELM mitigated}/H_{ELM natural}. It can be observed that the ELM mitigation is rarely realized when the SMBI is deposited at the bottom, especially outside of the pedestal. ELMs are mitigated with a tenfold ELM frequency than before when the deposition is almost located at the bottom to the center of the pedestal. Meanwhile, the energy loss and divertor heat flux are both reduced to one-third compared to the case of natural ELM. However, when SMB is further injected into deeper areas, the mitigation effect decreases rapidly. It can be seen that the ELM mitigation largely depends on the deposition of injected particles in the SMBI system. The suitable deposition of SMBI for ELM mitigation was 20% of the pedestal width inside the pedestal bottom on HL-2A.

Fig. 37

Deposition of SMB and its effect on ELMs. The electron density profile of edge plasma (black line), and the ratio of frequency, energy lost, and heat flux changed during SMBI (Yang et al. 2016)

5.2.3 Pedestal behavior with SMBI

The analysis of the parameter dependence sheds a light on the possible hidden physical processes owing to SMBI at the pedestal, where particle transport is crucial in ELM mitigation. Further analysis and observations of the pedestal behavior show a distinct possibility that the pedestal turbulence-related particle transport is responsible for ELM mitigation.

Figure 38 shows the difference in the density profile with and without SMBI on HL-2A. The density profile was determined using a microwave reflectometer. It can be observed that the density gradient (t = 701 ms) before SMBI is steeper than that of the case after SMBI (t = 722 ms). After a large ELM (t = 737 ms), the density profile was observed to recover, and the gradient increased to be larger than that before SMBI. A similar result was observed on the softened toroidal velocity profile, which was attributed to the damping effect of the deposited particles at the pedestal area (Xiao et al. 2012). The experimental observations suggest that enhanced transport events by SMBI degrade the pedestal particle confinement, which should also be responsible for the spanning of the pedestal span.

Fig. 38

The density profile evolution after the ELM mitigation with SMBI a SMBI valve signal, b D_α signal c density profiles with (722 ms) and without SMBI (701 and 737 ms) (Xiao et al. 2012)

The particle flux $\tilde{v}_{{\text{r}}} \times \tilde{n}_{{\text{f}}} /n_{0}$ is measured by the Langmuir probes. The change in particle transport was verified by the observed spectrum of the edge particle flux, as shown in Fig. 39. The low-frequency part of the particle flux spectrum decreased after the injection of SMB, whereas the high-frequency part increased. This result suggests that SMBI restrains the formation of large transport events (low-frequency), and induces smaller avalanches (high-frequency) (Gruzinov et al. 2002). This result also agrees with the decrease in the pedestal density gradient after SMBI, as shown in Fig. 38. Similar results were observed in the spectrum of edge density fluctuations measured by beam emission spectroscopy (BES) in KSTAR (Xiao et al. 2014b)

Fig. 39

Comparative results of the fluctuation-driven particle flux (Xiao et al. 2012)

Similar results have been observed for filament structures with SMBI injection in KSTAR (Kim et al. 2012). The filament structures in the ECEI images in Fig. 40 show that small filaments occur in large filaments. During the experiments, the SMBI was deposited inside the pedestal bottom. Apparent filament structures were observed, which varied significantly with the modification of the poloidal mode number of the filaments. The large filaments in Fig. 40a were later replaced by the small filament structures. The discrepancy in the filament structures in different periods might result from the modification of the ELM stability boundaries by SMBI.

Fig. 40

ECEI images of saturated edge filaments prior to the ELM crash (Kim et al. 2012)

More evidence from the analysis of the Langmuir probe data in HL-2A supports that SMBI could suppress the large-amplitude filaments, as shown in Fig. 41. The ELM mitigation lasted approximately 10 ms as shown by the shaded area. It was observed that the filament burst rate with a smaller amplitude threshold (δn_e/σ = 2) increased from 4 to 6 kHz and the larger amplitude threshold (δn_e/σ = 4) decreased from 2 to 1 kHz, where σ is the standard deviation of δn_e, and δn_e is the perturbation of n_e.

Fig. 41

The variation of the large filament events and small filament events with SMBI injection. a Divertor D_α signal, b, c filament burst rate with the threshold of 2 and 4 factors. d The link between filament burst rate and filament amplitude before (blue) and after (red) SMBI.(Nie et al. 2014)

In the divertor area, similar results were also observed on different scales of turbulence in the ELM mitigation experiments with SMBI as reported on EAST^. shown in Fig. 42. The divertor D_α signal in Fig. 42a clearly shows the ELM mitigation phase with SMBI. The particle flux ${\Gamma }_{Div}$ measured by the divertor Langmiur probes in Fig. 42b and its spectrum in Fig. 42c suggest that the intensity of large-scale turbulence decreases and the small-scale turbulence increases during ELM mitigation. The relationship between the particle flux induced by ELM events and the turbulence of different scale is shown in Fig. 42d, e. The particle flux released by the ELMs is represented by the symbol ${\Gamma }_{\mathrm{Div}}^{\mathrm{ELM}}$. ${\Gamma }_{\mathrm{Div}}^{\mathrm{ELM}}$ decreases with a decrease in the intensity of small-scale turbulence but grows with the enhancement of the large-scale turbulence. A model to interpret the ELM mitigation with SMBI has been proposed (Rhee et al. 2020). The simulation results show that SMBI enhances large transport avalanches caused by ballooning instabilities if the fueling is enough. They prevent the total pedestal current from reaching the boundary for peeling instability. Small-scale avalanches can be stimulated if the baseline fueling is low, preventing the pedestal from growing to profiles globally vulnerable to ballooning instabilities. More experimental investigations should be conducted to support the conclusion of the simulation.

Fig. 42

Time traces of a the divertor D_α signal and b the particle flux; c frequency power spectra of the turbulence, d particle flux ${\Gamma }_{Div}^{ELM}$ released by ELMs versus small (d) and large (e) scale turbulence intensity averaged over 40 ms (Zou et al. 2012)

5.2.4 Impurity SMBI

Helium (He) is the normal product in future fusion devices and is regarded as an impurity in current experimental devices. Generally, the injection of He SMBI often leads to an increase in the pedestal density gradient but a decrease in the ion temperature, especially at the edge. The pulse duration for the helium SMBI was usually shorter than that of the fueling SMBI to ensure good confinement of the plasma. From the ELM mitigation experiments with helium SMBI on HL-2A, it was observed that ELMs are mitigated by the He SMBI pulse as shown in Fig. 43. Meanwhile, it has been observed that the ratio of intensities of He-I (λ = 587.6 nm) and the D_α has a significant influence on the baseline of divertor D_α signal, indicating the increment of recycling gas (neutral gas). It should be observed that the T_i decreased up to 54% after the injection of He SMBI as shown in Fig. 43c. The ion temperature in the pedestal remained at a low level for around 80 ms owing to higher recycling after the injection of He SMBI. This indicates that the plasma confinement is degraded. Correspondingly, the H-factor decreases from 1.8 to 1.6 after helium injection.

Fig. 43

Time evolutions of plasma parameters for helium SMBI. (Ma et al. 2016)

The evolution of the ion temperature is shown in Fig. 44, where the intensity of He-II compared to its value before the He SMBI increased (> 1), and the T_i decreased. The modulation of the ion temperature by the He-II signal can be explained by the enhancement of recycling owing to He SMBI.

Fig. 44

Time evolutions of He II intensity (a), ion temperature (b), and D_α signal in divertor (c) (Ma et al. 2016).

A comparison of the pedestal parameters between the ELM mitigation with helium and deuterium was performed. The profiles of plasma parameters in the pedestal before and after He and D₂ SMBI are shown in Fig. 45. For both the deuterium and helium SMBI pulses, the slope of ion pressure in the pedestal becomes smaller after SMBI, and the well of E_r well is shallower because of the reduction in the ion diamagnetic term. The decrease in the pressure and its gradient by helium SMBI is less than that of deuterium simply because of the fuel quantity of the gas. Based on the above results, the cooling effect of SMBI plays a very important role in ELM mitigation, as it could decrease the ion temperature and its gradient in the pedestal. The T_i and its gradient had a significant impact on the radial electric field. Thus, the softened ion pressure induces a shallower E_r well, which may influence pedestal instabilities (see Fig. 46).

Fig. 45

Profile comparison between the helium and deuterium SMBIs. a electron density; b ion temperature; c ion pressure and (d) radial electric field (Ma et al. 2016)

Fig. 46

The dependence of ELM mitigation effect on the Ne impurity ratio in mixture SMBI

5.2.5 Impurity mixture SMBI

A mixture of the impurity gas and the fueling gas D₂ with different ratios seeded by the SMBI system, referred to as mixture SMBI, was first proposed and successfully mitigated the ELMs on HL-2A. The mixture gas-injection system was installed at the LFS of the HL-2A equatorial plane, which could modulate the gas pressure, pulse duration, and the ratio of the impurity gas to the fueling gas. Using this system, the effects of the SMBI mixture with different neon ratios (10%, 30%, and 100% Ne) on the pedestal instabilities were investigated in the typical H-mode discharge on HL-2Ashownas observed that the ELM behavior strongly depends on the ratio of the impurity component as shown in Typically, ELM mitigation is characterized by an increase of the ELM frequency and the decrease of the ELM amplitude. Thus, the frequency ratio before and after impurity mixture SMBI ${f}_{\mathrm{ELM}}^{\mathrm{after}}/{f}_{\mathrm{ELM}}^{\mathrm{before}}$ is used to present the variation of ELMs, where ${f}_{\mathrm{ELM}}^{\mathrm{after}}/{f}_{\mathrm{ELM}}^{\mathrm{before}}>1$ indicates ELM mitigation with an increase of the frequency. More data in Fig. 45 shows that the mitigation effect depends on the ratio of the impurity component, where 10% and 30% Ne in mixture SMBI can mitigate the ELMs. Observations such as an increase of the frequency and the ion temperature, suggest that the confinement is improved. The improvement of the confinement by the pure impurity gas seeded by the SMBI system is underway and further discussion is beyond the scope of this paper.

A typical result for ELM mitigation with a 10% Ne mixture SMBI is illustrated in Fig. 47. A moderate increase of the line-averaged electron density was observed after the 10% Ne mixture SMB. In addition, the total plasma radiation power in Fig. 47c also gradually increased. The global plasma confinement presented by the plasma-stored energy in Fig. 47e was only slightly degraded after the mixture SMBI. The pedestal density increases, as shown in Fig. 47g, whereas the plasma ion temperature at the edge in Fig. 47f is not influenced by the edge cooling effect and remains almost constant.

Fig. 47

ELM mitigation with mixture SMBI (10%Ne + 90% D₂) in HL-2A.Time evolution of a SMBI value signal, b line-averaged electron density, c total plasma radiation power, d divertor D_α signal, e core (blue) and edge (red) ion temperatures, f electron density at the pedestal top (Zhong et al. 2019)

Figure 48 shows the main parameters for the ELM mitigation experiment with 30% Ne mixture SMBI. As illustrated in Fig. 48(a), small-amplitude high-frequency bursts (HFBs) substitute the original ELMs as indicated by the spikes of the D_α signal. It has been observed that the baseline of the D_α signal increased with the HFBs, indicating the enhancement of pedestal particle transport. This result is also supported by the measurement of the pedestal density fluctuations, where the particle transport during HFBs is stronger than it during the inter-ELM phase. It was observed that the turbulence enhancement is mainly attributed to the enhancement of the high-frequency turbulence (> 50 kHz), which is modulated by the HFBs. From these observations, the pedestal turbulence enhancement might be responsible for the ELM mitigation.

Fig. 48

ELM mitigation with 30% Ne mixture SMBI in HL-2A. Time evolution of a Divertor D_α signal, b power spectrum of Doppler reflectometry, c integrated power spectrum for high (f > 50 kHz) and low (f < 50 kHz) frequency fluctuations, d electron temperature at the core and pedestal top, e electron density at the pedestal top, middle and foot (Zhong et al. 2019)

The divertor heat flux shown in Fig. 49 was obtained by Langmuir probes at the outer divertor plate near the striking point. As expected, the divertor heat flux peak significantly reduces, during which the heat flux induced by ELMs is replaced by that induced by HFBs with only 10% of the amplitude in Fig. 49(c).

Fig. 49

Response of divertor heat flux to 30% Ne mixture SMBI seeding. a divertor D_α signal, heat flux at the outer divertor plate b before and c after the SMBI (Zhong et al. 2019)

From the experimental results, plausible mechanisms of the ELM mitigation by SMBI with different gas species might be concluded. The injection of the SMBI modifies the peeling–ballooning (P–B) instabilities boundary or the location of the instabilities in the P–B boundary area, which directly changes the ELM behavior. This phenomenon is likely be observed when large quantities of particles are injected by SMBI strongly changing the local profiles and the parameters such as collision rate and confinement time as indicated in Sect. 5.2.2. In addition, it is commonly observed that the enhanced pedestal turbulence owing to SMBI fueling can affect the ELM behavior possibly through nonlinear interaction as introduced in Sects. 5.2.3 and 5.2.5. Especially for the high mass impurity gas, even the fueling rate is low. Other pedestal instabilities, such as high-frequency bursts observed in the 30% Ne mixture SMBI and some impurity-related modes, might be excited and affect the ELM behavior (Zhong et al. 2019). Besides, the E_r shear rate modified by SMBI fueling can also have a strong impact on ELM behavior as shown in Sect. 5.2.4, which is demonstrated by the simulation result (Zhu et al. 2022).

5.3 Active Control Of Divertor Heat With SMBI

SMBI has also been used to induce a radiative divertor on HL-2A with a mid-plane installed system (Yan et al. 2009). A completely detached plasma (CDP) discharge was obtained by the mid-plane SMBI, as shown in Fig. 50. It can be seen from the figure that the D_α emission in the divertor continuously decreases because of more ionization events in the upstream area. The comparison for the gas pressure in the divertor and main chamber is described by their ratio R_p0 = R_0d/R_0m, which is higher than 10, meaning much gas at the divertor area. The ionized component of the injected gas continuously contributes to the increase of the electron density with a maximum value of 4.6 × 10¹⁹ m⁻³. The injection of an impurity gas such as neon makes the situation much better. The density increased moderately compared to the deuterium injection. The experimental results show that it is difficult for the mid-plane SMBI to form a radiative divertor as well as maintain good confinement in the main chamber. The SMBI system in the divertor may provide a solution.

Fig. 50

A complete detachment plasma discharge with SMBI at the midplane (Yan et al. 2009)

Recently, the mid-plane SMBI was also valid in EAST radiative feedback control experiments, which improved the control accuracy and response speed(Yuan et al. 2020). This control system consists of a diagnostic (Langmuir probes), controller, and SMBI system, and can be applied to modify the radiative level under various conditions.

The SMBI can rapidly modify the density within 1–2 ms response time; however, the response time of the piezoelectric valve is slower (approximately 200 ms). As shown in Fig. 51, the peak density was set to 5.5 × 10¹⁹ m ⁻³. In the first two plots of Fig. 51, the ion saturation current density around the strike point in the upper outer divertor region and the divertor target temperature measured by infrared camera both indicate the reduction of the heat flux. In this experiment, mid-plane D₂ SMBI injection and the divertor impurity gas puffing are used, which could decrease the impact on the confinement in the main chamber. This result further indicates that the feedback control ability can be directly strengthened by the optimized actuator directly at the divertor, which is also highly desirable, especially for detachment maintenance during long-pulse operation.

Fig. 51

Active feedback control achieved by SMBI D₂ fueling at divertor region in EAST H-mode experiment (Yuan et al. 2020)

6 Summary

SMBI has considerable advantages in fueling and controlling particles after its first usage as the plasma fueling technique. Good directionality and high speed allow the gas jet injected by the SMBI system to penetrate deeper into the plasma than GP. Consequently, the fueling efficiency of SMBI is higher than that of GP, as demonstrated in many devices, indicating lower wall retention of neutral gas. Based on this characteristic, the SMBI system has also been used in transport studies. During the investigation of the particle transport, it is often operated in the modulated mode, which helps to determine the transport coefficient. In studies on heat transport, it is regarded as a cooling source, which can lead to a nonlocal phenomenon. In addition, the system has a fast response, which makes it a good candidate for density feedback control.

Owing to its deep penetration depth at the plasma edge, the SMBI technique is also extensively used in the H-mode-related studies, such as L–H transition and ELM control. It has been demonstrated that SMBI can trigger the L–H transition by lowering the H-mode power threshold via the nonlinear dynamics between the GAM and turbulence, which indicates that SMBI can be used as an external method to modify the L–H transition process under the condition of marginal heating power for H-mode. The ELM mitigation has also been successfully achieved with the fueling gas, impurity gases, and impurity mixture gas injected by the SMBI system. Several approaches to achieve ELM mitigation exist, such as the change in the instability boundary and the nonlinear interaction between the ELMs and pedestal turbulence. Regarding the impurity injection method, it becomes more complicated because the impurity has a significant influence on the instabilities and transport at the edge.

Owing to their flexibility, the quantity and species of the particles can be easily controlled. Thus, the SMBI technique is also treated as an active control method for the control of the heat load. Radiative divertors have been achieved on HL-2A and EAST, which significantly reduces the heat load on the divertor target.

In conclusion, SMBI is a useful tool for physical studies of fusion plasma in experimental devices. The validation of the high fueling efficiency on the experimental fusion devices indicates that this technique also has great potential on ITER when the pellet injection is not accessible such as in the case of helium plasma. To date, intensive work has been carried out to extend the capacity of the SMBI technique, especially by optimizing its directionality and speed with the newly designed integrated nozzle. Further efforts are still needed to focus on its use in fusion reactor environments with high magnetic fields, neutron flux, and far-field fueling issues.