Study on the L–H transition power threshold with RF heating and lithium-wall coating on EAST

The power threshold for low (L) to high (H) confinement mode transition achieved by radio-frequency (RF) heating and lithium-wall coating is investigated experimentally on EAST for two sets of walls: an all carbon wall (C) and molybdenum chamber and a carbon divertor (Mo/C). For both sets of walls, a minimum power threshold Pthr of ~0.6 MW was found when the EAST operates in a double null (DN) divertor configuration with intensive lithium-wall coating. When operating in upper single null (USN) or lower single null (LSN), the power threshold depends on the ion  ∇B drift direction. The low density dependence of the L–H power threshold, namely an increase below a minimum density, was identified in the Mo/C wall for the first time. For the C wall only the single-step L–H transition with limited injection power is observed whereas also the so-called dithering L–H transition is observed in the Mo/C wall. The dithering behaves distinctively in a USN, DN and LSN configuration, suggesting the divertor pumping capability is an important ingredient in this transition since the internal cryopump is located underneath the lower divertor. Depending on the chosen divertor configuration, the power across the separatrix Ploss increases with neutral density near the lower X-point in EAST with the Mo/C wall, consistent with previous results in the C wall (Xu et al 2011 Nucl. Fusion 51 072001). These findings suggest that the edge neutral density, the ion  ∇B drift as well as the divertor pumping capability play important roles in the L–H power threshold and transition behaviour.


Introduction
The high-confinement mode, i.e. H-mode, is the baseline operational scenario for ITER [1]. Since the power available in the initial phase of ITER operation is limited (~73 MW), accessing the H-mode with a heating power as low as possible is a crucial issue.
The capability of the flexible poloidal magnetic field control system in EAST can accommodate operation with different divertor configuration operations. In order to demonstrate high performance, long pulse plasma and in preparation for ITER, the carbon wall is being replaced by a metal wall step by step. The full carbon wall in the 2010 campaign (C wall) was replaced by molybdenum in the main chamber and carbon in the divertor in the 2012 campaign (Mo/C wall) [2]. EAST now routinely uses lithium coating as a main wall conditioning technique [3]. In addition, EAST has ITER-like heating schemes, i.e. dominated by electron heating from a lower hybrid current drive (LHCD) and ion cyclotron resonance heating (ICRH). These enabled capabilities make EAST a unique platform for the low/high (L-H) power threshold studies.
A theory-based prediction for the H-mode power threshold (P thr ) for ITER is still absent due to the ambiguous understanding of the physics of the L-H transition. The latest and widely used ITPA threshold scaling is P thr,08 = 0.0488 n e 0.72 B T 0.80 S 0.94 (MW), where n e is the central line-averaged electron density in 10 20 m −3 , B T is the magnetic field in T, and S is the plasma surface area in m 2 [4]. However, ongoing research from many devices found that other 'hidden variables' affect the L-H power threshold. The impact of divertor configuration on P thr is found in several devices, such as NSTX [5], JET [6] and EAST [7]. Recently, the L-H power threshold studies on JET [8] found that the P thr is lower when the PFCs are fully covered by metal. Motivated by these observations, the impacts of divertor geometry and lithium wall conditioning for two sets of wall materials on the L-H transition and P thr are studied in detail in EAST.
The paper is organized as follows: section 2 presents the experimental setup and a global description of the experiments. The main results in the previous L-H threshold experiments with the C wall in the 2010 experimental campaign are described in section 3. Section 4 addresses the findings obtained in 2012 with the Mo/C wall. A summary and discussion are presented in section 5.

Experimental setup
EAST is a medium size full-superconducting tokamak with a major radius R = 1.9 m and a minor radius a = 0.5 m. EAST has an ITER-like D shaped poloidal cross-section with a plasma area ~1 m 2 , plasma volume ~11 m 3 and plasma boundary surface area ~40 m 2 . From the initial operation phase up to 2010, EAST was equipped with a 2 MW LHCD and 1.5 MW ICRH systems. These two systems were upgraded to 4 MW and 6 MW respectively during its shutdown in 2011/2012. A detailed description of LHCD and ICRH experiments can be seen in [2]. During the 2010 and 2012 campaigns, EAST was equipped with one in-vessel cryopump which significantly enhanced the particle exhaust and capacity of the recycling control in H-mode discharges with a nominal pumping speed for deuterium of ~75.6 m 3 s −1 [2], which is located underneath the lower outer divertor target, as shown in figure 1. In order to avoid damaging the water-cooling pipes behind the pumping slots, the strike points on the divertor targets were shifted a few centimeters away from the slots in the 2012 campaign, relative to the 2010 campaign.
The discharges were conducted under the lithium-coated wall conditioning and detailed discussions of the lithium coating between the two campaigns can be found in Zuo et al's paper [3]. Although multiple L-H transitions can occur in a shot, only the first L-H transitions are studied in this paper. The L-H power threshold comparison is conducted by three different kinds of divertor configurations, i.e. lower single null (LSN), double null (DN) and upper single null (USN). In the 2010 campaign, the toroidal magnetic field, B T , and the plasma current, I p , were both in the anticlockwise direction viewed from the top (B × ∇B↑). In the 2012 campaign, there were few shots with B T in a clockwise direction (B × ∇B↓). The I p keeps the same direction. Note that, unfortunately, we have made a mistake with the field direction in the 2010 campaign (reference [9], the third paragraph on p 2). Here we have double checked the direction of B T and can deduce the field direction through the E × B flow velocity with the toroidal and poloidal rotation direction measured by the magnetic coils. In addition, the pinch angle of ELM filaments at the low field side gives us another approach to check the field direction. To make sure of this, the directions were directly measured utilizing a compass at multi-points at the vicinity of the EAST tokamak in the 2012 campaign. In this study, the L-H transitions are identified by the sudden drop of divertor Dα emission measured by a photodiode array and the increase of the line-averaged electron density n e measured by the central chord of the far infrared (FIR) interferometer. The lines of sight of the diagnostics are also shown in figure 1.
In this article, discharges were operated in deuterium with divertor configurations and all the L-H transitions in the plot occurred during the plasma flat top. The net power across the separatrix at the L-H transition, P loss , is calculated as, P loss = P aux + P ohmic − P rad − dW dia /dt, with P aux the absorbed auxiliary heating power from either one of the LHCD and ion cyclotron range of frequency (ICRF) heating, or combined power from both heating schemes (taking into account reflections and absorptions of the different heating schemes), P ohmic the Ohmic power, P rad the radiated power from the bulk plasma measured by absolute extreme ultraviolet (AXUV) arrays and dW dia /dt the change rate of the diamagnetic stored energy [9,10]. To improve the LHW coupling, the outer gap was optimized by isoflux plasma boundary control and local gas puffing near the LHW launcher. With these efforts the average reflection coefficient stayed low-in the range <10%. For LHCD, a combination of ray-tracing and Fokker-Planck calculations by using the C3PO/LUKE codes with so called spectral tail model are performed in order to identify the LH power deposition. It should be pointed out that the absorption coefficient in the experiment should be smaller than the theoretical value due to the spectral broadening. In the experiments, the absorbed fraction of the LH power can be determined with a method utilizing the time derivative of the total stored energy [11]. As the hydrogen concentration was reduced down to below 10%, effective ICRF heating was observed. The ICRF coupling efficiency can be simply estimated by calculating dW dia /dt at the time of the ICRF power turn-on [12]. The estimated error bars of P loss for the dataset are 10-20% of their own values. Most of the data have been obtained with a marginal input power in the C wall, which largely reduces the source of experimental uncertainties due to dW dia /dt. However, the dW dia /dt contribution might be larger due to the combined heating scheme of LHCD and ICRF in Mo/C wall studies. All global parameters are averaged over a time interval of 10 ms in the L-mode or intermediate phase just before the time of the L-H transition, where the impurity concentrations are low and the radio of radiation power to heating power is rather low. In this study, the L-H power threshold is studied in terms of two types of transitions, labeled as single-step L-H and dithering L-H transition, which can be identified by characterizing divertor Dα emission prior to the final transition. In the following, unless otherwise specified, small-amplitude limit cycle oscillations (LCOs) refer to the intermediate phase preceding the single-step L-H transition, and dithering cycles or the dithering phase represents the intermediate phase prior to the dithering L-H transition.

H-mode access and the typical L-H transitions
In the 2010 campaign, the first H-mode plasma appeared after strong lithium-wall conditioning both by lithium evaporation and real-time lithium powder injection at the plasma edge. The H-mode plasmas, typically with an H factor of H IPB98(y, 2) ~ 1, were obtained with ~ 1 MW lower hybrid wave power. In this campaign about 485 H-mode discharges were obtained. Stationary ELMy H-mode plasmas up to 6.4 s were produced in a wide range of operation parameters: B T = 1.4 -2 T, I p = 0.4 −0.8 MA, n e = (1.9 -3.4) × 10 19 m −3 , plasma surface area S = 38 -42 m 2 with DN (35% of H-mode shots) or intentionally unbalanced DN configurations (32% of H-mode shots) as well as LSN divertor configuration (13% of H-mode shots) with the ion ∇B drift towards the upper divertor. However, no H-mode was obtained with USN configuration. This is probably due to the limited heating power available then. In addition, there were 20% of L-H transitions that occurred during the plasma ramping phase, which is not discussed further here.
The small-amplitude LCOs prior to the L-H transitions firstly appeared in the 2010 campaign. These LCOs, characterized by small-amplitude (RMS/MEAN ~ 3%) and high frequency (up to 4 kHz) appearing in the target Dα emission signals with the input heating power very marginal to the trans ition threshold, were first demonstrated in [13]. The associated L-H transitions are single-step L-H transitions, which are typically characterized by a single-step reduction both in ion saturation current from the divertor target embedded Langmuir probes (figures 2(a) and (d)) and the divertor Dα emission (figures 2(b) and (e)). Some single-step L-H transitions are preceded by small-amplitude LCOs with a duration of cycles phase on ~100 ms time scale, with a frequency around 2 kHz, as shown in figure 2( f ). Note that these singlestep L-H transitions can be found both in DN and LSN configurations, with the ion ∇B drift away from the major X-point.

L-H power threshold and divertor configuration effects
All the L-H transitions were accessed by predominantly LHCD heating in 2010. In figure 3(a) the power across the separatrix P loss through the separatrix at L-H transitions are plotted versus the threshold powers P thr,08 predicted by the international tokamak scaling, showing that it basically follows the scaling in a limited region of operation windows. In figure 3(b) P loss are plotted as a function of n e All data available are in the low density range, n e = 1.9-3.4 × 10 19 m −3 . We did not see a clear density roll-over. This may be attributed to the very limited heating power which will be discussed in the following section.
Data has been sorted in terms of different divertor configurations to investigate its effects on the L-H power threshold (I p = 0.6 MA, B T = 1.56-1.78 T, n e = 1.9-3.2). In figure 4 P loss at the L-H transition are plotted versus P thr,08 with respect to different divertor configurations, showing that a minimum heating power of ~0.6 MW is needed for the H-mode access in the DN configuration, which is lower in comparison with that in the LSN configuration. Note that there are four points below 0.8 MW of the P loss for the LSN configuration, all of which were achieved after fresh wall conditioning (L-modes without transtion at a level of P loss ~ 0.8 MW can easily be found at similar plasma conditions). In addition, the H-mode cannot be accessed in the USN configuration with similar heating power, indicating a lower power threshold for the ion ∇B to drift away from the major X-point. This observation has been confirmed in a series of dedicated experiments (B × ∇B ↓) conducted in 2012 [7,14]. This is in contrast to other tokamaks [15][16][17][18][19][20].  In addition, all the type-I ELMy H-modes were obtained in the LSN configuration (B × ∇B ↑) in the 2012 campaign, suggesting that LSN with the ion ∇B drift away from the major X-point appears to facilitate achievement of highperformance stable H-mode plasmas on EAST. The internal cryopump is located underneath the lower divertor, which may help to reduce edge neutral density in the LSN configuration. On the other hand, a lower power threshold was observed in the USN configuration with B × ∇B ↓in the 2012 experiments [7,14,21], where the strike points on the divertor targets were shifted a few centimeters away from the pumping slots. Detailed discussions are made in section 4.2.

Role of edge neutral density on P thr
The importance of edge neutral particles or recycling to the L-H transition power threshold has been realized for a long time [22][23][24][25][26][27][28][29]. The neutral density is usually regarded as one of the most relevant 'hidden variables' behind the transition. The ion rotation and radial electric field E r at the plasma edge is supposed to be strongly dependent on charge-exchange momentum loss if the edge neutral density is high enough. Previous analysis [9] of the EAST H-mode results in 2010 with the C wall suggests that the neutral density near the lower X-point could be a key 'hidden variable' which affects the transition and power threshold. To access the H-mode on EAST, extensive lithium wall coating by evaporation was conducted every day before experimental operation. The neutral density in the edge plasma was estimated based on the divertor Dα emission measured by a photodiode array [30]. It was found that the neutral density near the lower X-point was reduced by a factor of four with heavy lithium wall coating, while the required minimum heating power to access the H-mode was gradually reduced from 0.6 MW to ~0.5 MW. Note that before the application of lithium coating no H-mode can be achieved.

H-mode access
In the 2012 campaign, EAST's capabilities have been significantly enhanced [2]. Graphite tiles on the low heat load area of the main chamber were replaced by molybdenum with the divertor remaining unchanged (Mo/C wall). In addition, the LHCD and ICRH systems were upgraded to 4 MW and 6 MW, respectively. The lithium evaporation system was upgraded to improve the coating uniformity, leading to a significant reduction in hydrogen concentration from ~10% down to ~3%, which allows for more effective ICRF heating with the minority heating scheme. With these enhanced capacities, H-modes with small-ELMs over 30 s have been achieved [2,31]. Note that before the application of lithium wall coating, no access to the H-mode can be achieved. In this campaign about 1371 H-mode discharges were obtained across a wide range of operation parameters: B T = 1.33-2.07 T,  H-modes were produced with different divertor configurations, including DN or unbalanced DN (75% of H-mode shots) as well as LSN (21% of the shots) and USN (4% of the shots).The L-H transitions in 16% of H-mode shots occurred during the configuration transient process from DN to LSN or USN configuration. There were 4% of L-H transitions that occurred during the plasma current ramping phase which are not included in the following analysis.  The oscillation amplitude (RMS/MEAN) for the single-step L-H transition is rather small, typically only of ~3% in target Dα signals and the divertor probe ion saturation signals, which is much smaller than that in the dithering L-H transition (~30%) (figure 6). These single-step L-H transitions are very similar to the L-H transitions which appeared in 2010, in which the oscillation amplitude sometimes slowly increased when approaching the transitions [13]. Significant magnetic perturbations, |δB P | ~ 1 G, have been detected by the Mirnov coils, associated with the 'dithering' or 'I-phase' in the dithering L-H transition which is normally obtained in the double null configuration [10]. The magnetic perturbations are much weaker when associated with the small-amplitude LCOs in the single-step L-H transition, which are normally obtained in the lower single null configuration. However, there is no significant difference in the L-H power threshold between these two types of transitions, indicating that the effect of rotating MHD on the L-H transition is not significant. On the other hand, recent experiments in HL-2A show that a kink-type MHD mode routinely occurs and crashes rapidly just prior to the I-H transition, which finally triggers the transition [32]. More experiments are required to investigate the effect of the rotating MHD on the L-H threshold power.
In figure 7(a) P loss are plotted versus P thr,08 for dithering L-H transitions and single-step L-H transitions, showing that the low boundary follows the scaling with P loss in a range of 0.6-1.7 times of P thr,08 . As shown in the figure, the loss powers P loss show no significant difference between the dithering L-H transitions and the single-step L-H transitions. However, the occurrence of these two types of L-H transitions strongly depends on the divertor configuration. Detailed experiments of the effects of divertor configuration on H-mode access are discussed in the following section. In figure 7(b) normalized powers P loss / P thr,08 are plotted as a function of n e . All the H-modes were operated in a low density range, n e = 1.4-3.6 × 10 19 m −3 . The low density dependence of the power threshold, namely an increase below a minimum density n e,min , was identified for the first time on EAST with the Mo/C wall. The minimum density n e,min is at about 2 × 10 19 m −3 , close to the minimum density limit for H-mode access in the 2010 campaign. This may be a reason why the density rollover was not observed afore-time. The physics mechanisms underlying the low density dependence of the power threshold remain largely unknown. Recent results from JET with a Be/W and C wall show a correlation of the density roll-over to divertor geometry and divertor/wall materials [8]. On the other hand, experiments from ASDEX Upgrade demonstrate the key role played by the edge ion heat flux as an explanation for the strong increase in the L-H power threshold at low density with electron cyclotron resonant heating (ECRH) as the auxiliary heating power [33]. These findings show that the occurrence of the low density dependence of the power threshold is not a universal feature in various tokamaks. More dedicated experiments in different heating schemes are required which will be available on EAST in 2015.

Influences of divertor configuration on the L-H power threshold and transition behaviour
EAST has a flexible poloidal magnetic field control system to accommodate different divertor configurations. Plasma configuration can change from one to another smoothly during single discharge. H-mode plasmas were obtained with different divertor configurations including LSN, USN and DN and different B T directions. Figure 8   Here, dR sep is the distance between the primary and secondary separatrixes at the outer midplane. Therefore, a value of dR sep = 0 indicates a perfectly balanced DN configuration while |dR sep | < 0.01 indicates an unbalanced DN configuration and a value of dR sep < −0.01 and dR sep > 0.01 indicates an LSN and USN configuration, respectively. As shown in figure 8(a), a minimum threshold value of P loss ~ 0.6 MW is found in the DN configuration while a lower power threshold is required in the LSN compared to the USN configuration (B × ∇B ↑). For B × ∇B ↓, the lowest threshold ~1.2 MW is also found in the DN configuration; both dithering L-H transitions and single-step L-H transitions are obtained in USN and DN, but cannot be accessed in the LSN configuration, which suggests that the power threshold is higher than the available power level with the ion ∇B drift towards the major X-point. To exclude the low density dependence of the L-H power threshold as mentioned before, the density range above 2.1 × 10 19 m −3 from the data set in figure 8(a) is selected and the normalized power threshold versus dR sep is shown in figure 8(b). It shows that there is a similar power threshold dependence for the ion ∇B drift direction towards the upper divertor, in which a minimum threshold is found in the DN configuration and a lower threshold is required in the LSN compared to the USN configuration. However, the dependence is not clear for the other side drift direction, where DN and USN share a similar power threshold. For DN, the power threshold is lower for B × ∇B ↑ when compared with that for B × ∇B ↓. Dedicated experiments are required to test the situation for B × ∇B ↓ which will be shown in the following paragraph. In addition, most dithering L-H transitions were obtained in the DN configuration for both B T directions, while almost all single-step L-H transitions were obtained in the LSN configuration for B × ∇B ↑ and in the USN configuration for B × ∇B ↓, which is normally pointing away from the major X-point for the ion ∇B drift on EAST. However, no significant difference of required heating power was found between dithering L-H transitions and single-step L-H transitions. This is further proved in a dedicated experiment with two adjacent shots running at the DN and LSN configuration with similar initial target plasmas, as illustrated in figure 9 (DN) and figure 10 (LSN). The dithering L-H transition and single-step L-H transition driven by ~0.5 MW LHCD and ~ 0.3 MW ICRH heating at a line-averaged electron density ~ 2.1 are obtained, respectively. To further investigate this, an experiment has been conducted in a single-shot with varying divertor configurations. As shown in figure 11, two L-H transitions were obtained in this shot with similar conditions while the configuration varied from DN to LSN (note that the density is even higher in the second L-H transition with the LSN configuration). The dithering L-H transition occurred in the DN configuration and the single-step transition appeared in the LSN configuration, which demonstrates the configuration dependence for dithering cycles preceding the L-H transition.
Not only the occurrence of the two types of L-H transitions strongly depends on the divertor configuration, but also the behaviour of the dithering cycles at the dithering L-H transitions significantly differ from one configuration to another. Figure 12 shows the target Dα signals of three typical shots for different divertor configurations, measured by an 18-channel vertical chord of Dα photodiode arrays (PDA) as shown in figure 1. The dithering cycles in the DN configuration are mostly regular large-amplitude oscillations, typically exhibiting a clear and sharp transition from the L-mode to the I-phase, as shown in figure 12(b). The duration of dithering cycles with the USN configuration (normally B × ∇B ↑) is usually very short, accompanied by only a few limits cycles ( figure 12(c)). The dithering cycles with LSN configuration (normally B × ∇B ↑) are more irregular in frequency and amplitude, and usually appear with increasing amplitude until the final transition to the ELMy-free H-mode ( figure 12(a)).   above are significantly reproducible for different divertor configurations. The magnetic configurations and main plasma parameters just prior to the L-H transitions of these three shots are shown in figure 13, with the plasma elongation factor κ, upper and lower triangularity, δ u and δ l , and the edge safety factor q 95 listed. The power across the separatrix P loss is also listed, indicating similar power injection for the three shots.
Although the appearance of the two types of L-H transitions shows no significant dependence on heating power, the duration of the dithering phase of the dithering L-H transition directly relates to it [34]. In EAST, dithering L-H transitions, dominantly heated by radio-frequency power, i.e. LHCD and/or ICRH, were selected to further investigate this issue (B T /I p ~ 1.8 T/0.4 MA, DN). We take the excess power divided by the confinement time τ E as an estimation of the ramp rate γ P of power in excess of the L-H transition power threshold at the separatrix, expressed as γ P = (P loss − P thr,08 ) /P loss /τ E . Figure 14 plots the duration of the dithering phase occurring at the L-H transition versus the ramp rate γ P . It shows that the duration of the dithering phase decreases with increasing γ P . To illustrate the effect, figure 15 shows three adjacent shots with similar plasma density ( figure 15(c)), and other parameters prior to the L-H transitions. The Dα signals have been shifted on the vertical axis to avoid overlap in the plot ( figure 15(a)). A progressive increase of the auxiliary RF heating power was conducted between these shots ( figure  15(b)), resulting in an increase in γ P . Due to this, a decrease of the duration of the dithering phase was found ( figure 15(a)). Meanwhile, the L-H transitions which occurred earlier correspond to higher power injection.
Dedicated experiments have been performed to further investigate the effects of divertor configuration on the L-H power threshold on EAST. Figure 16 shows three adjacent discharges with a similar initial target plasma density and input power from LHCD and ICRH under different divertor configurations for the ion ∇B drift towards the lower divertor. H-modes are achieved in the DN configuration (42023) with dR sep ~ 0 and the USN configuration (42024) with dR sep ~ 1.5 cm, as evidenced by the appearance of ELMs seen in the divertor Dα emission and the increase in line-averaged electron density. Note that the coupling power from LHCD is strongly reduced in DN after the L-H transition, while the H-mode can still be maintained for a long period at a higher density compared with the USN and LSN configuration. In contrast, the plasma remains in L-mode (shot 42022) for the LSN case with dR sep ~ −1.5 cm, suggesting that the power required for the L-H transition is lower with the ion ∇B drift away from the major X-point on EAST, which is in contrast to other tokamaks [15][16][17][18][19][20]. As mentioned above, EAST has only one in-vessel cryopump located underneath the lower divertor for providing the main pumping. However, the LSN configuration appears to have a higher power threshold. This indicates that the pumping capability is not so much sufficient to affect the L-H transition power threshold as the strike points on the divertor targets were shifted a few centimeters away from the pumping slots, as mentioned before.

Effects of lithium wall conditioning on P thr
To facilitate density control and reduce edge recycling for long-pulse operations, extensive efforts have been made in developing wall conditioning techniques [2,3,21]. In particular, we have explored various lithium coating techniques to enhance the uniformity of lithium coverage on the wall. In EAST, lithium wall conditioning has become routine and is the most effective wall conditioning technique to reduce neutral recycling. Compared with the 2010 campaign, similar techniques of lithium coating application, i.e. lithium  evaporation and lithium powder injection, were used in the 2012 campaign [3]. During the 2012 campaign, 15-45 g lithium was evaporated every day using three ovens. In addition, fine lithium powder was dropped in 47 H-mode shots. After shot 39637 (with ~0.765 kg of lithium accumulated) the EAST vacuum chamber had to be opened because of the breakdown of the fast control coil. Lithium coverage was cleaned up then. From the recovered experiment, strong lithium-wall coating was conducted with 1.613 kg of new lithium accumulated in total. To investigate the effects of lithium wall conditioning on the L-H transition threshold, H-modes with dithering L-H trans itions under similar conditions (I p = 0.28-0.6 MA, B T = 1.33-2.0 T, DN) are studied. We use the Dα emission near the lower X-point measured by a photodiode array divided by central line averaged electron density (Dα lower_X-point /n e ) as an indicator of the neutral density in the divertor region, as indicated in figure 1 (Dα (1-35a), red lines). The neutral density does not show an obvious downward trend when the lithium accumulation is less than 0.8 kg. However, the neutral density appears to gradually decrease after about 0.8 kg of lithium accumulation, as shown in figure 17(a). In addition, it is found that the power across the separatrix P loss increases with increasing neutral density while the power threshold P thr,08 is nearly unchanged with the similar initial plasma conditions (n e , B T , S), as indicated in figure 17(b). A reduced transition power threshold was observed with reduced edge neutral density, indicating that the edge neutral density near the X-point could be a key 'hidden variable' behind the H-mode access, which is consistent with previous experimental findings in EAST with the C wall [9]. Therefore, EAST experiments indicate that the low recycling regime with reduced edge neutral density enabled by lithium wall coating facilitates H-mode access, both for the C wall and Mo/C wall.

Role of field-dependent SOL parallel flow on the transition threshold
The ion ∇B drift direction affects the plasma flow along field lines in SOL. Measurements of the SOL parallel plasma flow (TCV [35], JET [36]) clearly indicate that there is a tendency for the plasma to flow along field lines in the co-current direction in SOL on the low field side near the midplane. These flows tend to reverse when the magnetic field is reversed. Similar results for the SOL flow reversal have been observed in the JT-60U tokamak by Mach probe measurements for two ion ∇B drift directions, suggesting that the ion Pfirsh-Schlüter (PS) flow is the most feasible mechanism to drive the parallel plasma flow at the midplane against the ion ∇B drift direction [37]. On EAST, the characteristics of measured SOL parallel plasma flow under various discharge conditions on the low field side midplane are consistent with the PS flow, with measured parallel Mach number M || = 0.3 − 0.5, as shown in figures 12 and 14 in [38,39]. These findings suggest that the SOL parallel flow on the low field side midplane is dominated by the PS flow and that the direction of the PS flow component of the SOL parallel flow depends on the ion ∇B drift direction.
To illustrate the configuration dependence of the transition threshold, a schematic diagram of the SOL parallel plasma flow in the LSN and USN configuration for two toroidal magn etic fields B T directions is shown in figure 18. The cases of counter-clockwise (B × ∇B ↑) and clockwise (B × ∇B ↓) directions of the field B T with the direction of plasma current I p remaining unchanged (counter-clockwise) are shown in figures 18(a)-(d ), respectively. As mentioned above, the L-H power threshold P thr is lower for the ion ∇B drift direction away from the major X-point than the opposite ion ∇B drift direction, i.e. a lower P thr was found in figure 18(a) than that in figure 18(b) and a lower P thr was observed in figure 18(d ) than that in figure 18(c), accordingly. A simple diagram of the SOL flow is illustrated: as driven by ballooning mode instability, a large part of the particles can cross the separatrix on the low field side due to unfavorable magnetic curvature there (indicated by red arrows), and then these particles move along the magnetic field line and reach divertor targets. The path connection lengths L || of the parallel flow in the SOL, which are dominated by the PS flow on the low field side midplane, are marked by black arrows for each case, with red circles showing the location of the divertor cryopump. The direction of the PS flow component of the SOL parallel flow depends on the ion ∇B drift direction. The velocity of the SOL parallel plasma flow V || at the outer midplane in EAST, which can be evaluated by the parallel Mach number M || (V || = M || * C s , with C s being the ion acoustic speed), is directed downwards (B × ∇B ↑) (figures 18(a) and (b)) and upwards (B × ∇B ↓) (figures 18(c) and (d )). The SOL parallel plasma flow direction at the outer midplane is reversed when the toroidal field changes to a clockwise direction with the same I p in a counterclockwise direction. When both B T and I p in the counter-clockwise direction are viewed from the top, the parallel plasma flow at the outer midplane SOL is directed downwards, i.e. towards the lower outer divertor in the LSN configuration and towards the upper inner divertor in the USN configuration, as shown in figures 18(a) and (b).     As shown in the figure, when the ion ∇B drift is directed away from the major X-point, the path connection length L || of the SOL parallel flow is much shorter than that with the ion ∇B drift directed to the major X-point. Therefore, the particle exhausts from the region near the separatrix at the outer midplane are transported to the divertor region in a much shorter timescale τ for the ion ∇B drift directed away from the major X-point due to the shorter path connection length L || of the SOL parallel flow, which can be evaluated as τ = L || / (M || * C s ). Once transported to the divertor region, the particles can effectively be screened and pumped out in the divertor region. This type of field direction-dependent particle transport pattern therefore benefits the particle exhaust. On the other hand, the particle exhaust with a long path connection length of the SOL parallel flow is not as sufficient as a shorter path connection length, since the time-extended cross-field transport in the SOL leads more particles beyond the divertor region which can hardly be pumped out. As mentioned before, the EAST results show that L-H transitions can be influenced by edge recycling and particle exhaust, and particle exhaust may depend on the path connection length of the parallel flow in the SOL. This may explain why the L-H power threshold is lower with the ion ∇B drift away from the major X-point. For the DN configuration, the path connection length of the SOL parallel flow is always short with respect to both the direction of the PS flow component of the SOL parallel flow and the opposite direction on the low field side midplane, despite the ion ∇B drift directions, which are probably coincident with the result of the minimum transition threshold achieved in the DN configuration. Also, the power threshold is lower for B × ∇B ↑ when compared with that for B × ∇B ↓ in the DN configuration, where the direction of the PS flow component of the SOL parallel flow is towards the lower divertor. A recent numerical approach of transition dynamics [40] shows that there is a strong nearly linear decrease in the total heat flux across the separatrix as the characteristic parallel advection loss time τ decreases. Generally, a short connection length makes the SOL more efficient and creates an ion pressure gradient across the SOL necessary to make the L-H transition, in agreement with experimental observations [19,41,42]. Thus, it suggests that field direction-dependent SOL parallel plasma flow boundary conditions may play an important role in the sensitivity of the L-H power threshold to configurations on EAST.

Summary and discussion
The L-H power threshold and transition behaviors have been studied with respect to divertor configuration and lithiumwall coating on EAST for two sets of walls: the C and Mo/C wall. Two types of L-H transitions, the single-step L-H transition and dithering L-H transition, have been identified. The dynamics of these L-H transitions have been discussed in [10] recently, exhibiting similar features of turbulence-flow interactions at the plasma edge evidenced by dedicated probe measurements. In the 2010 campaign with the C wall, the L-H transitions were mostly single-step L-H transitions with a heating power very marginal to the transition threshold predicted by the international tokamak scaling. Some single-step L-H transitions are preceded by small-amplitude LCOs with a frequency around 2 kHz (figure 2). However, both types of L-H transitions were observed in the 2012 campaign with the Mo/C wall, which can be accessed by a similar input power. On the other hand, a decrease of the duration of the dithering phase is found within the dithering L-H transition with increasing input power. These results suggest that the transition behaviour might be different even if input power is marginal to the transition threshold. A recent numerical approach of transition dynamics ( [43], and references therein), which demonstrated that LCOs appear when the input power is close to the transition threshold, should take more variations into consideration. This may not be the effect of the wall changes, since the wall material in the divertor was unchanged and heavy lithium wall coating was conducted in both campaigns. Another difference is that the strike points on the divertor targets were shifted a few centimeters away from the pumping slots in the whole campaign with the Mo/C wall. This shift significantly decreases the divertor pumping capability, compared with experiments with the C wall.
Results from the experiments with the Mo/C wall demonstrated that transition behaviour depends on the divertor configuration and the ion ∇B drift. For the ion ∇B drift towards the upper divertor, dithering L-H transitions were normally obtained in the DN configuration and almost all single-step L-H transitions were obtained in the LSN configuration ( figure 8). Dedicated experiments of two adjacent discharges with similar target plasma conditions also show that the dithering L-H transition occurred in the DN configuration and the single-step L-H transition was accessed in the LSN configuration (figures 9 and 10). This is further demonstrated by dedicated experiments varying configuration from the DN to LSN in a single-shot, in which the dithering L-H transition occurred in DN and the single-step L-H trans ition appeared in the LSN configuration (B × ∇B ↑) ( figure 11). For the ion ∇B drift towards the lower divertor, dithering L-H transitions were frequently observed also in DN, and single-step L-H transitions were normally obtained in the USN configuration, but not in the LSN configuration. In short, the dithering L-H transitions are usually accessed in the DN configuration with reduced divertor pumping capability, regardless of field directions. The single-step L-H transitions are normally obtained in a single null configuration with the ion ∇B drift away from the major X-point, no matter whether the divertor pumping is weak or strong. Thus, weak divertor pumping appears to be necessary for accessing the dithering L-H transition, whereas it does not appear to play a dominant role in setting the L-H transition behaviour. Similar divertor configuration effects are also observed on the Alcator C-Mod tokamak, where the dithering L-H transition is associated with a slot-divertor configuration and the single-step L-H transition is often found with a vertical-plate divertor configuration [44]. All these findings may suggest that there are some missing key ingredients, e.g. divertor configuration, ion ∇B drift direction, etc, in determining the transition behaviour. On the other hand, all the type-I ELMy H-modes were obtained in the LSN configuration (B × ∇B ↑) [14]. LSN with B × ∇B ↑ appears to facilitate the achievement of highperformance stable H-modes in EAST, as the cryopump is located underneath the lower outer target.
The divertor configuration dependence of the transition threshold is studied. For the C wall in the 2010 campaign, the H-mode can be accessed in the DN and LSN configuration for the ion ∇B drift towards the upper divertor, but is not accessible in the USN configuration with a marginal level injection power predominantly heated by LHCD. A minimum power threshold P thr of ~0.6 MW had been observed in the DN configuration ( figure 4). For the Mo/C wall in the 2012 campaign, L-H transitions had been accessed in different divertor configurations, i.e. LSN, DN and USN configuration, and B T directions with enhanced heating power. For the ion ∇B drift towards the upper divertor, a minimum power threshold P thr of ~0.6 MW is found in the DN configuration and a lower power threshold is required in the LSN compared to the USN configuration. For the ion ∇B drift towards the lower divertor, the lowest threshold ~1.1 MW is found also in the DN configuration. L-H transitions were obtained in USN and DN, but cannot be accessed in the LSN configuration ( figure 8(a)). For DN, the power threshold is lower for B × ∇B ↑ when compared with that for B × ∇B ↓. When the analysis is to convert from P loss to P loss /P thr,08 , a similar power threshold dependence can be seen for B × ∇B ↑ while the trend is not clear in the other side field direction ( figure 8(b)). For B × ∇B ↓, dedicated L-H transition experiments by varying configuration at similar initial target plasmas show that the L-H transition can be accessed in the DN and USN configuration, but is not accessible in the LSN configuration with the same heating power (figure 16). As mentioned above, EAST has only one invessel cryopump located underneath the lower outer divertor target for particle exhausting, however, the strike points on the divertor targets were shifted a few centimeters away from the pumping slots for the Mo/C wall. This results in a weak pumping capability which is not so much sufficient to affect the L-H transition power threshold. All of these findings demonstrate that a minimum P thr occurs in the DN configuration, and that a lower power threshold is required with the ion ∇B drift direction away from the major X-point for both the C wall and Mo/C wall on EAST, which is in contrast to other tokamaks [15][16][17][18][19][20]. Note that, in this study, the definition of divertor configuration in terms of small dR sep seems somewhat arbitrary and the presence of a second X-point inside the vacuum vessel and close to the plasma might affect on the edge plasma, wall interaction as well as recycling, which may result in the distinctive findings on EAST. On the other hand, similar results from ASDEX-Upgrade and MAST [19] and studies on NSTX [20] showed a minimum P thr also near balanced DN configuration. These findings show that the occurrence of a minimum in P thr is a universal feature in EAST, regardless of wall materials, pointing to a potential role played by SOL and divertor physics in the L-H transition. A schematic diagram of the SOL parallel plasma flow is discussed, which suggests that field direction-dependent SOL parallel plasma flow boundary conditions may play an important role in the sensitivity of the L-H power threshold to configurations.
Strong effects due to heavy lithium wall coating on the H-mode access have been observed both in experiments with the C wall and Mo/C wall. Similar results were obtained in the Mo/C wall, that neutral density near the lower X-point was progressively reduced by a factor of 2 with increasing lithium accumulation, and that the power across the separatrix P loss decreases with increasing lithium accumulation (figure 17), compared with previous studies in the C wall [9]. The low edge recycling conditions achieved by lithium wall coating is thought to be the key factor behind the low L-H power threshold, as indicated by a significant drop in the neutral density near the lower X-point. All these findings may suggest that low recycling conditions achieved by lithium wall coating facilitate H-mode access on EAST, probably including ITER. The low density dependence of L-H power threshold P thr , namely an increase of P thr below a minimum density n e,min , is identified in the Mo/C wall for the first time, not in the C wall. The minimum density n e,min is close to the minimum density limit for H-mode access with marginal input power in the C wall. This may explain why the low density dependence is not seen in the C wall. Future works are required to test the low density dependence of the power threshold in different heating schemes, such as neutral beam injection heating. In this study, the difference in P loss for different configurations is not so significant only when we consider the source heating power, and the possibility within the difference in absorbed power that is related to the divertor configuration and ion ∇B drift direction needs more investigation. Future works are required to accumulate dedicated experiments and diagnostics, which will be available in the next campaign in 2015.