Experimental study of electron cyclotron heating assisted start-up on J-TEXT

Second harmonic X-mode (X2 mode) electron cyclotron heating (ECH) assisted start-up has been studied experimentally on the J-TEXT device to determine the minimum ECH power requirements to assist breakdown and develop a better physical description of the process. Results indicate that the minimum toroidal electric field for a successful start-up on J-TEXT changed from 2.5Vm−1 to 0.56Vm−1 , and that injecting 300kW of X2-mode ECH power can ensure robust breakdown. The critical ECH power for successful start-up was determined to be approximately 200kW . At lower powers ECH was observed to cause ionization, but this did not necessarily result in successful start-up. The effects of varying ECH power, pulse-width and toroidal magnetic field on start-up were also investigated. Higher ECH power leads to quicker, stronger ionization and CIII emission, and is beneficial for burn-through. Higher pre-plasma density caused by high ECH power can decrease the required toroidal electric field for ohmic breakdown, while the enhanced CIII emission may not be good for start-up. The characteristics of the pre-plasma formed by ECH prior to application of the loop voltage were also studied. The toroidal magnetic field affects the initial location where pre-plasma forms and this also affected the subsequent tokamak start-up. The analysis of spatial density showed that the pre-plasma can radially develop at a velocity of 600ms−1 . Furthermore, it was found that injecting ECH power at the appropriate time with a low power of 150kW can achieve similar pre-ionization results to the high-power case. The transition from ECH plasma to ohmic plasma suggests that the ECH assisted start-up modified the process of purely ohmic breakdown.


Introduction
Plasma initiation is the first step in a tokamak discharge. It is usually achieved by an inductive toroidal electric field (E loop ). E loop for superconducting tokamaks such as JT-60SA [1] and the International Thermonuclear Experimental Reactor (ITER) [2] is limited due to the restrictions on the voltage that can be applied over the superconducting central solenoid. Thus, such devices can only initiate plasma at E loop ⩽ 0.4 V m −1 . Electron cyclotron heating (ECH) assisted start-up can relax the requirement on E loop by forming an initial plasma (or pre-plasma) before the application of E loop (pre-ionization) [3] and providing additional power during burn-through [4], making it a promising method for assisted start-up and accepted as a scheme for ITER first plasma [2].
The uncertainty of a significantly low E loop ( ∼0.3 V m −1 ) for ITER start-up requires further experimental research to strengthen knowledge on tokamak initiation, and especially how ECH can assist in this process. The classic zerodimensional simulation estimated the ECH power requirements for ITER [4]. Earlier ECH assisted start-up experiments at DIII-D confirmed that a low E loop start-up with ECH assistance proposed for ITER is acceptable, and there are many advantages in ECH assisted start-up [5]. Experiments conducted in tokamaks such as JT-60U [6], DIII-D [7], Tore Supra [8], Korea Superconducting Tokamak Advanced Research (KSTAR) [9] and Frascati Tokamak Upgrade (FTU) [10] also verified the feasibility of low E loop start-up with ECH assistance. These studies are concerned with the ECH, the prefill pressure, the magnetic field and the polarization. Some other items are also important, such as experimental simulation [11], multi-machine comparison [12] and trapped particle configuration [13]. These experiments provided a scientific basis for ECH assisted ITER start-up. However, the detailed physical process of the ECH pre-ionization, the formation and sustainment of the pre-plasma and the required ECH power to ensure effective breakdown assistance in ITER is not sufficiently clear, meaning it cannot be confidently modeled [14].
ECH assisted start-up experiments have been conducted at J-TEXT since 2019. These experiments aim to determine the minimum requirements for ECH to assist in the breakdown and develop a better physical description of the pre-ionization process and transition from ECH pre-plasma to ohmic plasma. They can improve our theoretical understanding and contribute to model development. The rest of this paper is organized as follows. Section 2 describes the experimental setup. The analysis of the poloidal magnetic field configuration and experimental results, such as typical discharge and the effect of ECH power and pulse-width on start-up, are provided in section 3. Section 4 discusses the ECH threshold power for ionization and makes a comparison with other devices. A summary is given in the last section.

Experimental set-up
J-TEXT is a conventional medium-sized iron-core tokamak with a major radius R 0 of 1.05 m. It operates at a minor radius a of 0.25-0.29 m with a movable titanium carbide coated graphite limiter. A typical J-TEXT discharge in the limiter configuration is done with a toroidal field B T of 2.0 T, a plasma current flat-top I p of 200 kA, a pulse length of 800 ms, plasma densities n e of (1-7) ×10 19 m −3 and an electron temperature T e of 1 keV [15].
For easy understanding of the following text, we briefly introduce the J-TEXT discharge. The discharge action of J-TEXT is as follows [16]. As J-TEXT has an iron core and discharges with the capacitors, its discharge differs from many large running tokamaks. Figure 1 shows the circuit structure diagram of the J-TEXT ohmic heating system and its working form. Firstly, the pre-magnetic rectifier discharges to the ohmic poloidal field coils, generating a reversed current. Then the pre-magnet capacitor discharges the pre-magnetized rectifier, forcing the latter to switch off. The timing when the pre-magnetic capacitor is triggered is artificially defined as t = 0.0 ms. Note there exists a delay of up to 1 ms in the discharge. Then the ionization capacitor begins discharging to the ohmic coils immediately. The hysteresis effect and rapid current growth in the ohmic poloidal field coils induces a large flux change in the tokamak center, resulting in an E loop in the toroidal vacuum vessel. This starts the process of plasma initiation. For J-TEXT, the other four parallel capacitors named C 1 to C 4 (not shown in figure 1) work in sequence to discharge to the ohmic poloidal field coils, forming the current ramp-up, flat-top and ramp-down process. A successful start-up means that plasma current can reach the pre-set flat-top value, usually 120 kA, 160 kA or 180 kA. However, as we do not change the discharge condition of capacitors C 1 , C 2 , C 3 and C 4 , a lower plasma current of 40 kA can be used to determine the success or failure of start-up.
The distribution of related diagnostics and the ECH system on J-TEXT in top view is shown in figure 2. Key diagnostics include a multichannel three-wave far-infrared polarimeterinterferometer (named 'POLARIS') [17,18] and a photodiode array (PDA). POLARIS has 17 channels spaced 3 cm apart along the radial direction. Its measurement error is usually lower than 0.01 × 10 19 m −2 . Two kinds of PDA are used in this experiments. The PDA applied in the earlier experiments, corresponding to most shots in this paper, has a radial measurement channel at the low-field side. The shots with a number greater than #1068946 in this paper used a new PDA, which has 14 measurement channels. All the measurement channels are perpendicular to the radial direction and spaced by 2 cm, allowing the PDA to measure H α from R − R 0 = −12 cm to R − R 0 = +16 cm, where R and R 0 are the radial position in the cylindrical coordinate system and the plasma major radius, respectively.
In 2019 an ECH system was developed at J-TEXT that can apply high-powered radiofrequency waves with a frequency of 105 GHZ and a maximum injected power of 500 kW for a maximum duration of 1 s, which enabled these breakdown experiments. Its launcher is installed at the low-field side of the middle plane port. The nearly independent drive mechanism and the steering mirror allow it to inject ECH power into the vessel with a toroidal injection angle of ±20 • and a poloidal injection angle of ±20 • [19]. In this work, the ECH  power was injected into the vessel with a poloidal angle of +3 • and a toroidal injection angle of 0 • . This small poloidally inclined angle can avoid direct reflection at the high-field side back into the ECH launcher. Two polarizers installed at two miter bends can achieve arbitrary polarization, allowing injected ECH power at the second-harmonic X mode (X2 mode) in all shots presented in this paper. Note that for the low-density plasmas considered during these plasma initiation experiments, only a small fraction of the X2 mode ECH power is absorbed. This paper quotes injected ECH powers, not absorbed powers. A small ECH power contributes less to the pre-ionization. Hence, the maximum temporal ECH power before the inductive discharge is more important than the averaged ECH power.

Experimental results
The poloidal magnetic configuration can affect the start-up result, which is introduced here first. Typical ECH assisted startup can be obtained after optimizing the injected ECH power and the ohmic electric field, i.e. the E loop induced by the central solenoid. Experiments have been conducted by applying ECH power prior to the time the E loop is induced. Hence, the ECH power is used not only to support the burn-through of a plasma created by the breakdown due to the application of E loop but also to pre-ionize the pre-fill gas, creating a so-called preplasma, prior to the application of E loop .We study details about the ECH pre-ionization by scanning the ECH power and its pulse duration, showing its effect on pre-plasma formation as well as the following tokamak plasma formation by the application of E loop . Furthermore, the evolution of spatial density can be investigated to learn more about the pre-plasma behavior and how it in turn can affect the tokamak start-up process. The following sub-sections give a more detailed description. The poloidal magnetic field configuration (in gauss) in cross-section when t = 0 ms [22]. The microwave path (red dotted line) and the resonance layer are marked. Reproduced from [22]. CC BY 4.0.

Poloidal magnetic field configuration
Initial electrons ionized from the neutral gas will be lost to the wall along magnetic field lines. A longer field line benefits the development of plasma initiation. As the poloidal magnetic field can decrease this distance, the analysis and optimization of the poloidal magnetic field become important.
The evolution of poloidal field coil currents can be seen in figure 1. Based on the EFUND code and related discharge parameters, the distribution of the poloidal magnetic field can be estimated. Figure 3 shows the spatial distribution of the magnetic field configuration in cross-section when t = 0.0 s as was used in these experiments. The ECH resonance layer locates around the center of the vessel (namely R − R 0 = 0 cm), as the toroidal magnetic field (TF) is 1.87 T in all these experimental shots. It can be observed from figure 3 that the poloidal field around the resonance layer is about 3 G. The toroidal connection length L = 1590 m can be estimated from equation (1) in [4], where the minor radius a = 0.255 m, the TF B ⊥ = 1.87 T and the poloidal stray magnetic field B ⊥ = 3 G at the core. Such a distance is enough to achieve a successful breakdown in a large pre-fill range without the assistance of ECH [4]. The average connection length the electrons may experience on their path along the torus to the wall can, however, be lower than this, as they move radially into areas that have a higher stray magnetic field. Giving B ⊥ a larger value of 10 G, one would find a plasma current of 1.3 kA as the value where one would get closed flux surfaces [14] L ≈ aB T /B ⊥ . (1) It is known that providing a magnetic null improves ohmic breakdown, due to the longer connection length. On the other hand, ECH assisted breakdown experiments have shown that a small vertical magnetic field can be beneficial to the preionization process [12,20,21]. In these experiments, the above magnetic null configuration is used throughout to simply allow a better comparison between ohmic and ECH assisted plasma initiation attempts without changing any other parameters.

Typical ECH assisted start-up
It is well known that ECH can expand the operational space by relaxing the requirement of E loop for start-up, especially in the breakdown process. The minimum toroidal electric field (E min ) for breakdown can be determined by changing the voltage of the ionization capacitor, without ECH preionization. Injecting ECH power prior to E loop , we can observe the effect of pre-plasma on E loop . Many shots are used to explore the electric field limitation. Figure 4 shows a statistical relationship between the injected ECH power and E loop . More details can be found in the caption to figure 4. Figure 4 contains four major items of information.
(i) The E min for a successful ohmic breakdown is about 1.925 V m −1 . If E loop is below such a value, plasma current cannot exceed 40 kA during start-up. Effect of the injected ECH power on the E loop required to create a tokamak discharge. The ECH power is injected prior to the application of E loop , thus this effect is not due to burn-through assistance. Note that the voltage of the ionization capacitor is also modified to investigate the limitation of E loop for pure ohmic heating start-up and ECH assisted start-up. The first number in the legends denotes the voltage of the ionization capacitor. Characters D1 and D2 mean the experiment was carried out on different days. The indication 'Fail' means that plasma current did not manage to exceed 40 kA during the startup.
(ii) ECH assisted start-up expands the operational space. Pure ohmic heating start-up failed at an E loop of 1.6 V m −1 . However, ECH assisted start-up succeeds in a start-up with a much lower loop voltage of 0.56 V m −1 , which is the E min achieved by ECH assisted start-up on J-TEXT. When an ECH power of about 200 kW is injected, the start-up is robust. It should be noted that this ECH power is the peak value before the inductive discharge. This will be further studied in the later part of this sub-section. (iii) The critical ECH power for start-up is between 90 kW and 180 kW. As there is little difference in E loop in two shots (marked 'A' and 'B'), the effect of ECH power is weak, indicating the critical ECH power for a successful startup is close to 180 kW.  Thus the results show that the preplasma created by the injection of ECH somehow helps with the subsequent formation of the tokamak discharge, allowing start-up at a lower E loop . However, the decrease in the electric field is not obvious if the ECH power is larger than this critical value of 250 kW. In other words, when the ECH power is greater than 250 kW the reduction of E loop is not obvious.
The temporal evolution of related signals show further details of the start-up process. Figure 5 shows three typical shots, including the conventional ohmic heating start-up (#1068920), a successful ohmic heating start-up with E min (#1065724) and a typical ECH assisted start-up (#1068924). For a robust start-up, J-TEXT usually discharges with an extremely high breakdown loop voltage of about 34 V (E loop = 5.2 V m −1 , not presented in figure 4), which can be observed from the loop voltage peak in figure 5. This E loop is much higher than the required E loop (0.3 V m −1 ) for ITER. The E min for a successful ohmic breakdown is 2.5 V m −1 . An E min of 0.56 V m −1 can be achieved by injecting ECH power from −17 ms to 10 ms. The voltage of the ionization capacitor is set to 1 V so that the flux provided by the ohmic coils almost reaches its lower limit. Under such a condition, the magnetic hysteresis effect could affect the electric field. Another possible reason may be related to the wall conditioning. There is no special wall conditioning apart from Taylor-type discharge cleaning and baking at 150 • C, which makes it possible to gather more impurities from the vessel. Although this E loop is higher than 0.3 V m −1 , we can conclude that ECH power can relax the required E loop for the breakdown and expand the operational space. The initial plasma (pre-plasma) is a key characteristic in ECH assisted start-up. H α , pre-plasma density and CIII emission begin to increase before the inductive discharge, as shown in figure 5. The appearance of H α and pre-plasma density means the existence of pre-plasma, and thus part of the prefilled gas is already ionized before the application of E loop . When CIII emission is detected, it suggests that there are electrons in the plasma with an energy of about 30 eV. The emission is optimum if the average electron temperature is about 30 eV. Thus if the CIII emission starts to decrease it would suggest that the average electron temperature is increasing above this value. Based on POLARIS, the spatial density distribution can be measured to explore the development of preplasma. Figure 6 shows the spatial density distribution of preplasma density at three key moments of shot #1068924. The pre-plasma forms at about t = −2 ms. At that time, the preplasma density is below the noise level of the diagnostic. Soon after, t = −1 ms, a significant signal is observed. The second moment (t = 1.6 ms) is before the application of a reversed E loop . The third moment (t = 3.1 ms) is when the ionization capacitor is applied.
At t = −1 ms, the density peak is at the center of the vessel, where the ECH resonance layer is located. This phenomenon means that pre-plasma generates (or forms) near the resonance layer, as observed in many tokamaks [10,23,24]. At t = 1.6 ms, the plasma density becomes radially asymmetric. Pre-plasma tends to move to the low-field side, although the pre-plasma density also increases at the high-field side. This phenomenon can be understood by drift motion and ⃗ E × ⃗ B convection. At t = 3.1 ms, there is nearly no growth in plasma density at the high-field side, while apparent density growth occurs at the low-field side. However, the edge electron density on both sides decreases, causing a peak density profile. Compared with the second moment, the center of plasma density also moves towards the center of the vessel. This phenomenon should be related to the enhancement of the poloidal magnetic field, especially the vertical field and the ohmic field. Comparing this shot with the other two shots shown in figure 5, the pure ohmic heating start-up still does not have initial electrons, while shot #1068924 has a central line-integrated density of about 0.035 × 10 19 m −2 . As the loop voltage (or E loop ) consists of the sum of the plasma resistive component (I p R p ) and the inductive component (L c dI p /dt), where I p is the plasma current, R p the plasma resistance and L c the plasma inductance [14,25,26], these electrons can enhance the conductivity of the pre-fill gas (hydrogen) but contribute less to the reduction of the inductive component, relaxing the requirement of loop voltage for the avalanche process during the ohmic heating breakdown. As the E min of ECH assisted start-up is 29% of the ohmic heating start-up, the consumption in volt seconds by ECH assisted start-up is 16% of that of ohmic heating start-up until t = 10 ms. The cost is that the plasma current by ECH assisted start-up is 36% of that of ohmic heating start-up.

ECH power and pre-ionization
The typical discharge suggests that injecting ECH power before the start of the inductive discharge can achieve preionization. However, the relationship between the production of initial electrons and ECH power is still unclear. A further experiment focusing on exploring the effect of different ECH powers on pre-ionization, especially the pre-plasma density, was carried out on J-TEXT. Figure 7 shows the temporal evolution of several related signals when different ECH powers are injected. The discharge conditions for these shots are as follows. The ECH powers applied to shots #1073817, #1073819, #1073820 and #1073825 are 365 kW, 230 kW, 300 kW and 390 kW, respectively. In these shots, the common factors, such as prefill pressure and vertical field, which may affect the start-up, are set to the same for the purpose of signal variables. The voltage of the ionization capacitor is set to 1 V. Hence, the hysteresis effect, rather than the ohmic field caused by the currents in the center solenoid, should contribute more to E loop during the first 10 ms in the inductive discharge.
Except for shot #1073820, it can be observed that a higher ECH power corresponds to an earlier appearance of H α and pre-plasma density. In other words, a higher ECH power can achieve quicker ionization in a certain ECH power range. H α increases earlier with higher power but also reaches a maximum earlier. The minimum delay between the onset of ECH power and the appearance of H α is about 15 ms. The delay from the appearance of H α emission to its peak can be as short as 4 ms with nearly 400 kW ECH power. In the later related experiments on J-TEXT, the former delay can be as short as about 2 ms. The delay is 3 ms on DIII-D [24] and about 23 ms on KSTAR [9]. This delay can be further shortened in J-TEXT by injecting a higher ECH power and optimizing parameters such as the vertical field and pre-fill pressure.
When H α reaches a maximum, the pre-plasma is thought to be predominantly ionized, and the degree of ionization might increase after this time to even higher values. Figure 6 shows that such H α rollover is obtained for the two cases with the highest ECH powers (i.e. #1073825, #1073817) while this has not yet been achieved for the other cases with lower powers. However, only the appearance of ionization cannot support a successful start-up, as shot #1073819 shows. A high-quality pre-plasma means that the pre-plasma density and temperature effectively support tokamak start-up. A higher ECH power contributes a higher quality pre-plasma to support tokamak start-up. This quality is linked to the degree of ionization that is achieved (i.e. a higher pre-plasma density) and the spatial extension of the pre-plasma. Small and lowdensity pre-plasmas are not as effective.
A further comparison of these shots shows that a high ECH power can achieve a strong ionization and a higher density after ECH power is applied for enough time. Comparing the H α between the shot with the minimum ECH power (#1073819) and the shot with the maximum ECH power (#1073825), one can find that their amplitude characterizations before and after the application of E loop are opposite. Shot #1073825 can achieve a strong H α before the application of E loop . Its H α amplitude caused by the ECH power alone is about two or three times larger than that caused by the inductive discharge and ECH power together. On the contrary, for shot #1073819, the H α amplitude before the application of the E loop is two times smaller than that caused by the ohmic power and ECH power together. Besides that, for shot #1073825, the H α amplitude caused by ECH power alone is similar to that of shot #1073819 caused by ECH power and ohmic heating together. These phenomena indicate that the ECH power can achieve as strong ionization in a local area as the ohmic heating power. Based on the distribution of H α (shot #1073825), we can give the bold conjecture or prediction that the pre-plasma density will not increase much under such a configuration, even if the ECH power is increased.
The reason for the failed start-up of shot #1073819 can be understood as follows. As there is no loop voltage peak at about t = 3 ms (ohmic breakdown often happens at about t = 3 ms; note shot #1068920 in figure 5) and the plasma current appears at about 10 kA at t = 10 ms, this shot succeeds in breakdown. The failed start-up of shot #1073819 is caused by the quality of pre-plasma or insufficient power to support tokamak plasma start-up. The large difference in the temporal evolution of H α before and after t = 0 s can verify this. If we put capacitor C 1 into use at t = 7 ms or inject high ECH power, shot #1073819 will possibly achieve a successful start-up.
As pre-plasma density is important, we analyze the spatial density of four shots shown in figure 7 before the application of E loop (t = 1 ms), as shown in figure 8. We can find that a lower ECH power corresponds to a low number of particles in a section profile. This should be one of the reasons why shot #1073825 succeeded in start-up while shot #1073819 failed. Figure 9 shows the density profile at t = 2 ms and t = 8 ms for shots #1073819 and #1073820. It can be easily observed that these two shots have a similar electron density at t = 2 ms. However, the density profile is obviously different at t = 8 ms. The peak density of shot #1073819 is located at R − R 0 = −21 cm, while that of shot #1073820 is at R − R 0 = −3 cm. The differences in profiles at this stage of the discharge are most likely determined by the input power. As the ohmic power is the same for these discharges, it is thought that the difference is due to the variation in the ECH power used .

ECH pulse duration and pre-ionization
The evolution of pre-plasma density depends on the injected ECH power. In this section, the dependence of the duration of the ECH (at constant power) on the pre-plasma and the tokamak start-up is investigated. Figure 10 shows the temporal evolution of several key preplasma and tokamak discharge start-up parameters when the injected ECH power is started at different times, thus changing the ECH pulse duration. It can be observed that a short ECH pulse with lower power of 150 kW (#1068924) can achieve a similarly successful start-up to those shots with high ECH power parameters. This means that the pre-plasma formation time, i.e. the time needed to form a pre-plasma with sufficient quality to affect the tokamak start-up, depends on the ECH power, but when lower powers are applied for longer the same effect can be achieved. For shot #1068928, the time from the start of the ECH injection to the appearance of a clear H α signal is 19.3 ms, while for shot #1068924 it is 14.9 ms. Thus, high ECH power does not solely determine the pre-plasma quality and it should be applied for a sufficiently long duration to be effective as well. It is not entirely clear what physics sets the minimum duration for a given ECH power, but it is possible that it is linked to the processes that allow the radial expansion of the pre-plasma.
Hence, the profile of pre-plasma before the application of E loop was investigated to explore the possibility of particle spatial distribution. Figure 11 shows the line-integrated density profiles for the four cases shown in figure 10, taken before the application of E loop . All profiles are radially asymmetric with respect to the ECH resonance position. This is typical for the formation of ECH plasmas with open-field lines [7,9,10,24] and will be discussed further in section 3.6. Although the input ECH power is low, shot #1068924 has a rather large preplasma density, which makes its start-up successful. The preplasma density development for this shot has leveled off from about −1 ms to 1 ms, i.e. the density is constant in time. This phenomenon suggests that too long an ECH pulse does not necessarily mean the density continues to increase. As mentioned before, this is probably set by the available particles in the system (i.e. in first order those provided by the pre-fill) and the pre-plasma particle and energy balance. The earlier application of ECH, as is shown for shot #1068928, allows the density to increase earlier but it levels off at a similar density towards the end of the ECH pulse (i.e. from t = −1 ms to 1 ms) When pre-plasma density reaches its balance density, the input ECH power will maintain the existing pre-plasma in a dynamic rather than increasing pre-plasma density.

Toroidal magnetic field
The TF affects the ratio of the vertical field to the TF and the location of the ECH resonance layer, leading to the change in connection length, as pointed out by equation (1). In this section, the effect of TF on pre-ionization is investigated. Figure 12 shows the temporal evolution of several key signals of five shots with different TF. It can be observed that when the TF is located at the high-field side (R res = −9.8 cm and R res = −4.2 cm), the onset time of H α is about 5 ms earlier than when the TF is located at the center (R res = 0 cm) and about 9 ms earlier than when TF is located at the lowfield side (R res = 3.6 cm). R res is the radial location of the resonance layer with respect to the core. It can be calculated by R res = R − R 0 . When B T = 2 T, there is no visible rise in H α emission or in pre-plasma density. Measurement of the spatial density suggests that there is no pre-plasma density rise in all the chords, indicating no pre-ionization. The subsequent growth in pre-plasma density in the 5 ms after the appearance of pre-plasma indicates that pre-plasma created at the high-field side tends to increase about two or three times quicker than plasma created at the core and about five to seven times quicker than plasma created at the low-field side (R res = 3.6 cm). The pre-plasma density growth rate can also be observed. Five milliseconds after pre-plasma appears, the pre-plasma density growth rate at the high-field side is about two or three times larger than that at the core and about five to seven times larger than that at the low-field side. The density amplitudes have a similar quantitative relationship. The peak pre-plasma density has a similar quantitative relationship to the pre-plasma density rise ratio. All these phenomena indicate that pre-ionization at the high-field side tends to create preplasma earlier, which develops quicker and achieves a higher peak pre-plasma density.
These phenomena could be related to the connection length. As shown in equation (1), the connection length is associated with the radius, TF and the poloidal magnetic field. Figure 3 illustrates that the low-field side tends to have a lower poloidal magnetic field of 2.4 G. When pre-ionization occurs at the low-field side, the TF at the core is higher, meaning that B T /B ⊥ increases when TF increases. However, the effective minor radius, which is defined as the radial distance from where pre-plasma is created to the wall, decreases as TF increases, indicating there may exist an inflection point in experiments, although the calculation indicates the effective connection roughly increases from 1412 m to 1752 m. This is counterintuitive, like the result of scanning a vertical field on DIII-D [27]. The possibility of inaccurate estimation exists while the electromagnetic environment in experiments is very sophisticated. As the electromagnetic environment in experiments is sophisticated, there exists the possibility that the estimation of the vertical field is not accurate enough. One should note that a small vertical field of 0.25%-0.5% of B T can obtain a higher pre-plasma density for wall conditioning on the TCV tokamak [28] (although in our experiments, we changed the toroidal field rather than the vertical field). Figure 13 shows the spatial line-integrated density distribution and the location of the resonance layer. It can be Figure 11. The spatial density distribution of four shots before the application of E loop (t = 1.3 ms). For J-TEXT, its major radius and minor radius are R 0 = 1.05 m and a = 0.255 m, respectively, meaning that the location where R − R 0 = 0 cm is the center of the vessel, while the locations R − R 0 = −25 cm and R − R 0 = 25 cm are close to the inner wall (high-field side) and the outside wall (low-field side), respectively. The ECH resonance layer is located in the center of the vessel. easily observed that the location of the resonance layer is close to where the peak density is located, indicating that preionization starts near the resonance layer Figure 14 shows the spatial density distribution at t = 6 ms. It can be observed that for the case of a high TF, plasma density can increase drastically and form a peaked plasma density profile. The case of a low TF also has a rather peaked plasma density profile. Note that the density center of the former locates at the low-field side while the density center of the latter locates at the high-field side. This phenomenon suggests that the different TFs change the location of plasma initiation, leading to different density profiles at the moment of plasma current initiation. For tokamak operation, a higher or lower TF can be taken into consideration. However, an extremely high TF causes failed plasma initiation, which may be related to a rather short connection length because a high TF can apparently decrease the effective radius. A sharp density increase can frequently be observed once the loop voltage is applied to a low pre-plasma density, as shown in shots #1068936 and #1068923. However, the reason for such a phenomenon is still unclear. Figure 15 shows the spatial density evolution from t = −30 ms to t = 10 ms at different TFs. When pre-ionization occurs at the high-field side, the formation of pre-plasma is close to t = −10 ms. Subsequently, pre-plasma moves from the high-field side to the low-field side during pre-ionization. However, plasma develops from the high-field side to the inner wall. The subsequent tokamak start-up occurs at the high-field side. When pre-ionization occurs at the core, preionization becomes weak. However, the subsequent tokamak start-up appears earlier, and the ionization is stronger than all the other cases shown in figure 15. Plasma develops from the low-field side to the high-field side. When preionization occurs at the low-field side, pre-ionization becomes weaker than all the other cases. The subsequent tokamak start-up begins near the resonance layer and expands to both sides. These phenomena indicate that the location of the preplasma affects the location where tokamak plasma begins to form. The mechanism may be that the existence of preplasma makes the breakdown of neutral gas easier, which makes the creation of tokamak plasma over a larger region possible.

Spatial density distribution during the pre-ionization and the inductive discharge
The spatial evolution of line-integrated density at several key moments is investigated to explore the plasma behavior during pre-ionization and the inductive discharge. The different behaviors of pre-plasma and plasma can be observed. It may also teach us how the pre-plasma is converted into a current carrying plasma after E loop is applied. Figure 16 shows the spatial evolution of line-integrated density at several time slices for shot #1073825. At a time early in the development of the pre-plasma (about t = −10 ms), the plasma density is low at only a few locations, just larger than the measurement error. The behavior of pre-plasma can be described as follows. Firstly, pre-plasma forms near and around the resonance layer, at R − R 0 = −3 cm from about t = −10 ms to −9 ms. Then from about t = −6 ms, the growth of pre-plasma density at the core stalls. This phenomenon is related to the plasma obtaining a balance in parallel and perpendicular particle transport. The position of peak density shifts with respect to the ECH resonance from R − R 0 = −3 cm to R − R 0 = 0 cm. Pre-plasma expands on both the high (R < R 0 ) and low-field (R > R 0 ) sides, but as is typical for such a kind of plasma it is asymmetric, with higher densities found on the low-field side. More details about the development of these asymmetric pre-plasma profiles will be discussed later.
In time, the pre-plasma density profile continues to broaden, or to expand radially, always with the tendency of density growth inclined to the low-field side. A rough comparison between the density distribution at t = −2 ms and t = 1 ms finds that the total particles at these two moments are nearly equal. This indicates a balance of particle number.
At the moment t = 3.6 ms, E loop has been applied for 2 ms and reaches nearly its maximum value. Plasma current has reached 4 kA, indicating that close flux may already form at this point, significantly improving the plasma confinement. The density distribution is now more uniform. Besides that, plasma density reduces to three-quarters of the size it achieved at t = 1 ms. Many particles are likely lost to the wall due to the quick change in the poloidal magnetic field configuration. The density drop after the application of an E loop can also be found in recent work on DIII-D [24] and ADITYA-U [29].
It can be observed that there is a steady increase in all radial positions from t = 3.6 ms to t = 5 ms. The density increase around R − R 0 = −3 cm is most apparent. As plasma current builds up, a closed surface begins to form. The electron density at the low-field side decreases while the electron density at the high-field side increases. Particles at the low-field side may    move to the high-field side, or the ionization at the high-field side becomes strong.
As the plasma current keeps increasing, the radial density distribution becomes more uniform, i.e. less asymmetric. Also, the density begins ramping up again, indicating that plasma tends to move along a fixed toroidal direction, and a closed flux surface begins to form. The spatial density evolution before and after the application of E loop , and thus for the ECH generated pre-plasma and during the ohmic tokamak discharge, is different.
We now look further into the asymmetry of pre-plasma density. At t = −6 ms, the pre-plasma density on the highfield side and on the low-field side is about 5.5 × 10 16 m −1 and 9.0 × 10 16 m −1 , respectively. This estimation is based on the assumption that plasma density is uniform in a halfwidth of 1.5 cm of both sides of measurement chord. In other words, if the line-integrated density at R − R 0 = X cm is N, we assume that the line-integrated density in the area R − R 0 = [X − 1.5 cm, X + 1.5 cm] is also N. The calculation result shows that the particle number on the low-field side is nearly twice as large as that of the high-field side. This difference is bigger if toroidicity is taken into consideration.
The asymmetry development of pre-plasma density can be semi-quantitatively explained as follows. The pre-plasma diffuses radially outward from the ECH resonance. But, besides a diffusive component, it is thought that in these plasmas the transport also has a strong convective component, due to the possible build-up of radial electric field. Such fields could form due to the opposite vertical drift motion of ions and electrons in an open-field line system.
Comparing figure 16 with figure 8, we find that the spatial density distributions of four shots at t = 1 ms are very like the spatial distribution of shot #1073825 at several different moments. This phenomenon indicates that low-power ECH start-up has a similar evolutionary process, although it requires more time to reach a similar spatial distribution. However, limited by the local wave field, the low ECH assisted start-up cannot reach the same peak electron density as the case assisted by a high ECH power.
Note that the plasma current is different at t = 15 ms in figure 12. We further analyze it from the perspective of spatial density distribution. Figure 17 shows the evolution of spatial density at t = 15 ms when scanning the TF. It can be observed that for shots #1068920 and #1068926 its central plasma density is a little higher while its plasma current is higher and loop voltage is smaller, indicating that a central plasma may benefit rising plasma current and a low toroidal electric field.

Pre-plasma radial expansion velocity
Based on POLARIS, we can observe the radial pre-plasma expansion from generation to the edge. In this sub-section we estimate the radial pre-plasma expansion velocity.
The multi-channel density behavior of shot #1073825 is shown in figure 18 We do not show the line-integrated density at R − R 0 = 6 cm because its signal is corrupted. Four points marked with letters connected by a red line can be seen in this figure, indicating the time when pre-plasma begins to increase for this channel. The point indicated by the letter A also refers to the moment of appearance of H α , as shown in figure 7.
The detailed radial expansion of pre-plasma is shown in figure 18. At first, the pre-plasma stays in a local area for about 2 ms after generation. In this process, the growth of pre-plasma density is like an exponential function. Then the pre-plasma expands rapidly towards the low-field side (point B to point C) at a high speed of about 600 m s −1 in the low-field-side direction. Closer to the edge, its radial speed is slower (about 125 m s −1 ) As mentioned before, diffusion, drift and convection are the possible factors affecting the radial expansion (towards the low-field side) speed. As there is no closed flux surface before the application of E loop , parallel transport can also be a key factor affecting the movement of particles. A rough evaluation in the Appendix suggests that the radial expansion speed of 600 m s −1 cannot be explained by the Bohm diffusion and parallel transport.
The remaining possibility is the convection caused by ⃗ E × ⃗ B drifts due to the build-up of large vertical electric fields in such a type of open-field-line plasma. As this is a much more sophisticated process it is not easy to calculate. Simulation by code such as TOMATOR-1D [20] may be able to provide an answer to this question. We also pay attention to the relationship between the radial expansion rate and ECH power. For shot #1073819 with a low ECH power, the pre-plasma radial expansion velocity is very low and the density ramp-up also becomes unremarkable. More shots are required for further research.

Transition process from pre-plasma to tokamak plasma
The next step is to investigate how the pre-plasmas generated by ECH can affect the subsequent ohmic tokamak discharge start-up. Figure 19 shows the start-up temporal evolution of two shots. The first case is a pure ohmic heating start-up (#1065728). Note there is a 2 ms delay between the appearance of peak voltage and the formation of plasma at the high-field side. The plasma initiation develops from the high-field to the low-field side. The radial plasma expansion velocity from the high-field side to the core is approximately 200 m s −1 . The second case (#1065738) is assisted by ECH, starting well before the application of E loop , at t = −27 ms, creating a pre-plasma. As known, from the discussion in the previous sections, the pre-plasma is formed around the center of the vessel at where the ECH resonance layer is located and an asymmetric radial pre-plasma density distribution develops before the application of the toroidal electric field. When the revised toroidal electric is applied, the pre-plasma density at the low-field side decreases rapidly. When the tokamak plasma forms, with closed-flux surfaces, the main particle type lost to the vessel wall changes from ions to electrons as indicated   by the measured vessel current at the low-field side. When the normal toroidal electric field is applied (about at t = 3 ms), the case with an ECH generated pre-plasma, shows formation of plasma current about 1.5 ms earlier and an increase in plasma density that is twice as fast as for the case without the preplasma.
It should be noted that the central pre-plasma contributes nearly fully to the later inductive discharge while the preplasma at the low-field side is lost to the wall. Pre-plasma makes the process of breakdown happen over a large range near the center of the vessel at the same time, indicating that the processes of purely ohmic breakdown and ECH assisted breakdown are different. For ECH assisted start-up, plasma can be formed earlier (about 1.5 ms) because the toroidal low-density plasma eliminates the delay of ohmic breakdown (including the avalanche process).

ECH threshold power for pre-ionization and comparison with other devices
Electron density, rather than H α , can intuitively represent the level of ionization, assuming it is a fraction of electrons that were present in the pre-fill pressure prior to the application of ECH. Figure 20 illustrates the relationship between the maximum pre-plasma density and the ECH power, showing a nonlinear relationship. The minimum ECH power needed to be able to achieve ionization can be estimated as of the order of 10 kW from [30]. Under such a minimum power, ionization by ECH is impossible. Figure 20 suggests that this value is for J-TEXT below 220 kW. If one takes shot #1068924 into consideration, this value should be below 150 kW. An accurate value requires more shots. For ECH powers above 300 kW, the pre-plasma density increases faster and to higher values faster. The tendency may be like an exponential function. However, for ECH powers above 400 kW, the pre-plasma density is not expected to increase further because the balance between ionization and loss quickly settles. The exact value at which this would happen is still unknown because of the limited available ECH power and experimental time.
We may conclude the following possible relationships between injected ECH power and pre-plasma density maximum, as shown in figure 21.
• When ECH power is smaller than P 1 , no ionization happens.
This value can be predicted by Farina's equation. As shown before, for J-TEXT this value is of the order of 10 kW. • When ECH power is between P 1 and P 2 , electron density will increase if higher ECH powers are applied. However, the average ionization of the pre-plasma might still be weak.  The possible ECH power threshold P 2 is like a critical electric field that determines whether the avalanche-like process can occur.
• When the injected ECH power is between P 2 and P 3 , electron density will increase noticeably. This is an avalanchelike process. When ECH power is larger, electrons can gain enough energy from the wave field so that electrons can collide with the neutrals and generate further ionization. • When the injected ECH power is larger than P 3 , electron density increases a little, even when the injected ECH power increases a lot. The major reason for this is the limitation of the available gas in the considered system (i.e. the J-TEXT vessel).
There exists the possibility that there are only two thresholds (without P 2 ). As mentioned above, the ECH threshold power for creating ionization in J-TEXT is 150-200 kW. One could estimate the EC pre-plasma volume to scale roughly as R 0 × a ∼ R 2 0 (where R 0 is the device major radius and a the minor radius) [24]. If we want to achieve a similar condition in DIII-D (R 0 = 1.67 m), the required ECH power is estimated to be 0.38-0.5 MW, which is close to the result of 0.5 MW on DIII-D [24]. The required ECH power to create ionization in ITER (R 0 = 6.2 m) is roughly estimated to be 5.2-7 MW. Note that this threshold power is based on the application of X2 mode ECH. As the electron density and temperature are low, the ECH beam possibly passes through the resonance layer several times due to reflection.

Summary
ECH assisted start-up experiments have been conducted on J-TEXT. These experiments focused on pre-ionization by ECH, and thus the creation of a pre-plasma, prior to the onset of the toroidal electric field and thus the standard tokamak discharge initiation. A minimum ECH power for successful creation of a pre-plasma, that effectively assists the initiation of tokamak discharge in J-TEXT, was found to be about 150 kW. Some common pre-ionization phenomena such as H α development and its rollover and a typical asymmetric pre-plasma density (profile) development have been observed.
Higher ECH power leads to quicker ionization, higher preplasma density, stronger CIII emission and eventually a burnthrough, i.e. an H α rollover, during the pre-plasma stage. The relationship between ECH power and maximum pre-plasma density may be a piecewise function. When ECH power is lower than a critical value, which is theoretically predicted by [30], the ECH will not be able to ionize and thus create a pre-plasma. However, more power is needed to obtain a high-quality pre-plasma than just for ionizing. When ECH power is lower than 300 kW in J-TEXT, the growth rate of preplasma density is slow. But for higher powers (300-400 kW) the pre-plasma density can be increased more quickly. The spatial evolution of pre-plasma was explored to examine the behavior of pre-plasma and ohmic heating plasma. POLARIS provides the radial development of pre-plasma. The physics that determines the quick radial expansion velocity of 600 m s −1 is unclear.
A short ECH pulse with low ECH power can achieve a successful start-up as effectively as shots with high power and a long pulse. Enough ECH power is required, and a long pulse-width is unnecessary. Turning off ECH power too early is not conducive to burn-through.
The scan of TF shows that the location of ECH resonance at the high-field side rather than the low-field side improves its effect on pre-plasma formation. The location of the pre-plasma affects where the tokamak plasma begins to form.
We further discussed the ECH power thresholds and multimachine comparison. A power threshold to create ionization, an avalanche-like plasma formation and reach a density balance is likely to exist from experimental observations or theoretical derivation. The threshold power to create ionization for ITER is roughly estimated to be 5.2-7 MW, which can be scaled from J-TEXT and is supported by the experimental result on DIII-D.
Many parameters, such as the delay (between the onset of ECH power and the appearance of pre-plasma), and some critical powers, can be affected by factors such as magnetic field configuration and pre-fill pressure. The tendency should be the same when these factors are changed unless the change is so large that start-up fails. Many shots are expected to provide more data. In addition, comparing experimental data with tokamaks of different sizes can offer more data, verify the repeatability of some phenomena and support multi-machine research.
is far lower than the experimentally estimated value, and thus the observed expansion speed cannot be explained by Bohm diffusion.
Then we consider the possibility of parallel transport. For an open-field configuration, a similar transport loss can be considered as a transonic ambipolar flow along a magnetic field line towards the vessel wall [32]. The hydrogen ion sound speed can be estimated by where m H is the hydrogen ion mass, 1.66 × 10 −27 kg. For a pre-plasma, the ion energy can be considered negligible if compared with electron energy, assumed again to be 30 eV.
The ion sound speed can be estimated as 5.4 × 10 4 m s −1 . The time when plasma moves to the edge can be estimated by [32] τ = L C s ≈ 0.03 s.
This is much longer than the required expansion time seen in figure 18, which is approximately 0.32 ms. The possibility of a combined effect can also be excluded due to the large difference in the order.