Investigation of performance enhancement by balanced double-null shaping in KSTAR

We report experimental observations on the effect of plasma boundary shaping towards balanced double-null (DN) configuration on the plasma performance in KSTAR. The transition from a single-null to a DN configuration resulted in improved plasma performance, manifested through changes in the pedestal region, decreased density, and core MHD activity variation. Specifically, the DN transition led to a wider and higher pedestal structure, accompanied by grassy edge-localized modes (ELMs) characteristics. The density decrease was a prerequisite for performance enhancement during DN shaping, increasing fast ion confinement. Optimizing the plasma near the core region was associated with the suppression of sawtooth instabilities and the occurrence of fishbone modes during the DN transition. Integrated modeling demonstrated that secondary effects of the DN shaping could increase core thermal energy confinement.


Introduction
Controlling the plasma shape is one of the available options for optimizing the tokamak plasma operation together with different operational control knobs (such as plasma current I P , toroidal magnetic field B T , external heating, and gas injection).It is possible to control the shape near the plasma boundary by adjusting a series of poloidal field coil currents.Magnetic diagnostic systems and equilibrium reconstruction techniques such as EFIT [1] allow us to guess the magnetic geometry inside the tokamak during operations.Due to its good accessibility for real-time control, plasma shaping has become an experimental research topic for studies.For instance, in recent studies using machine learning methods, the plasma shape was an excellent actuator [2,3] and controlled target [4].
Furthermore, it has received much attention theoretically with the consideration of the real tokamak geometry in the last few decades.In the magnetohydrodynamics (MHD) stability, plasma elongation κ and triangularity δ play a critical role in the β limit [5] and the internal kink mode [6][7][8].The Hmode edge pedestal prediction based on the theory of coupled peeling-ballooning mode (PBM) [9,10] also showed that the strong shaping could substantially expand the stability boundary.This expansion can change the pedestal structure and type I ELM onset [11][12][13][14][15][16][17][18].Moreover, the X-point effect on edge stability has a stabilizing effect, specifically on the peeling modes [19,20].
In agreement with the above studies, the energy confinement scaling laws have shown meaningfully positive weight on the elongation.According to ITER-89 [31], an L-mode energy confinement time scaling, and IPB98(y,2) [32], an ELMy H-mode thermal energy confinement time scaling, the exponent of elongation is 0.50 and 0.76, respectively.Both have the third largest value after major radius and plasma current.Due to the influence of the PBM stabilization effect discussed, it can be confirmed that elongation is more significant in H-mode.Aside from the results of the multi-machine database regarding κ, previous comprehensive studies using specific sets of experiments have reported that strong shaping, including triangularity and even squareness ζ, has favourable effects on confinement enhancements in both core and pedestal regions in various devices, DIII-D [33][34][35], JT-60U [36][37][38], AUG [39], and JET [40].
It is important to explore the appropriate shape for numerous scenarios and approaches to effectively generate fusion energy, satisfying operational conditions; highperformance, steady-state, and long-pulse.To meet those challenges, advanced scenarios, such as hybrid and steady-state scenarios with internal transport barriers, have been proposed as potential candidates for ITER [41,42].For steady-state operations, exploring self-generated current sources is underway other than bootstrap current [43][44][45][46].
At the same time, there is another practical constraint limiting operating conditions.Excessive power and particle loads from edge localized modes (ELMs) in H-mode plasma must be well-controlled because they can substantially damage plasma-facing components in future high-power tokamak reactors [47].Several solutions have been developed, including active ELM suppression by resonant magnetic perturbations [48,49], naturally ELM-free [50,51], and even a high-performance L-mode regime [52].There are also many efforts to minimize wall loads by control outside the plasma, for example, detachment [53,54] and various magnetic divertor configuration designs [55,56].
The null configuration, considered one of the plasma shaping options, could be attempted to satisfy the afore mentioned requirements.Previous studies about double-null (DN) configuration with upper and lower X-points, which will be mainly discussed in this paper, have generally focused on the edge and scrape-off layer (SOL) phenomena [57][58][59][60][61][62].Due to divertor heat flux sharing [63], DN shaping is a viable approach if enhanced performance is guaranteed.However, there are few experimental studies utilizing this configuration for performance enhancement, except the high-performance DN experiments in DIII-D [64,65].These studies focused on the performance enhancement by deuterium gas puffing on the DN plasmas.Our work investigates the experimental results of KSTAR discharges with performance enhancement during the single-null to DN transition.In these results, altering the null configuration was an active trigger of the transition to a highperformance regime, not a passive background.
At the outset, we clarify that only the connected or balanced DN configurations are considered as DN in this paper.KSTAR has the shaping capability to allow plasmas with elongation up to 1.9, triangularity up to 0.85, and vertically symmetric DN configurations [66].The unbalanced DN magnetic configuration, mainly downward-biased lower single-null (LSN, favourable ion grad-B drift direction for the L-H transition), is generally used in KSTAR to obtain H-mode plasma easily.
In a dedicated high-performance transition discharge, we will observe the aspect of transition, and the emphasis will be on how it changes the entire plasma due to the DN shaping.However, the discovery of new mechanisms during performance enhancement is beyond the scope of the present paper.Therefore, we will limit ourselves to outlining and analyzing the experimental results based on the existing theories.Also, note that performance enhancement is not always observed in the KSTAR DN experiments.
This paper is organized as follows.In section 2, we present the experimental observations from the KSTAR DN experiments, including the method, results of performance enhancement, observations on the pedestal, the changes of density, and the core MHD activities.The detailed analysis of the performance enhancement due to DN transition are discussed in section 3, addressing pedestal stability, enhancements in the fast ion channel, and providing a comprehensive insight through integrated modeling.Finally, our conclusions are drawn in section 4.

Method of KSTAR DN shaping experiments
The DN shaping experiments have been designed to compare the LSN and the DN plasma in a discharge of the KSTAR hybrid scenario [67].We used a feedforward control of dRsep within one single discharge to change the null configuration.The definition of dRsep is the radial difference between the upper and the lower separatrix flux surface at the outboard mid-plane.The DN shaping control began in the stationary phase of an H-mode LSN plasma with dRsep of −2 cm to −1 cm (the negative value of dRsep means LSN).Then, the shape evolved to DN until dRsep reach almost 0 cm. Figure 1(a) shows the plasma boundary shape variation from magnetic equilibrium reconstruction using EFIT [1] during the DN transition.By the shaping control, the upper Xpoint was symmetrically connected with the lower X-point.The DN configuration was achieved without changing the mid-plane and lower strike points at both the inboard and outboard.In figure 1(b), dRsep gradually increased from 7.3 s (black dashed vertical line) to 8.7 s (green dashed vertical line).As the elongation increased from 1.7 to 1.8, the upper triangularity doubled from 0.42 to 0.84.

Results of performance enhancement
Figure 2 is an example of a discharge with performance enhancement during the DN transition, shot 25460.Edge safety factor at 95% of last closed flux surface (LCFS), q 95 increased along with the dRsep variation.Normalized beta, β N increased from 2.1 to 2.7 and the confinement enhancement factor, H 89 rapidly increased from 2.0 up to 2.5 after 8.4 s (magenta dashed vertical line) when the dRsep was around −0.3 cm, order of the ion Larmor radius, ρ i in KSTAR.Internal inductance, l i showed a little change about 5% around 1.0.The D α intensity measured from the mid-plane decreased and showed a grassier ELM burst after the transition was almost finished.The line averaged density gradually decreased as it approached DN.On the contrary, the electron temperature from electron cyclotron emission (ECE) measurements [68] increased with the suppression of the sawtooth-like behaviors during the DN transition.The neutron emission rate from the neutron detector increased with the transition.This increase in the neutron emission rate indicates an enhancement in the fast ion channel, which is the primary trigger for the higher fusion reaction rates in this neutral beam injection (NBI) heated discharge.The discharge was sustained until the discharge ended at 10 s as pre-programmed.It should be noted that the synchronization of each change during the sequence of this experiment was subtly decoupled.When the dRsep reaches 0 at 8.7 s (green dashed vertical line), the line-averaged density stopped decreasing and remained constant.However, β N and H 89 started increasing rapidly from 8.4 s (magenta dashed vertical line), and it continued to increase even from 8.7 s to 9.0 s (red dashed vertical line).The dramatic increases were observed simultaneously in temperature and neutron rate.At 8.4 s, where to start the rapid increase, there was a significant change in the ELM characteristics, as shown in the D α signal.Before 8.4 s, the giant ELM bursts were regular; after that, they mixed with small ELMs.from Thomson scattering (TS) [69] are plotted in figures 3(a) and (b), respectively.Note that the radially different five chords of the two-color interferometry (TCI) measurements [70] were considered together in n e fitting with TS in the profile reconstruction.The ion temperature, T i , and the toroidal rotation velocity, V tor , measured from charge exchange spectroscopy (CES) [71] are presented in figures 3(c) and (d), respectively.The radial coordinate is the normalized toroidal flux coordinates, ρ N , from the kinetic EFIT reconstruction.
In the LSN configuration from 7.3 s to 8.0 s, κ and δ did not increase significantly.By this time, the density entirely decreased while only a few amounts of temperature increases were observed in the edge region.Before the completion of the DN configuration at 8.7 s, temperatures showed remarkable increases in both core and pedestal regions, while there was a continuous drop in density.Even after the boundary shape was fixed in the DN configuration and the density decrease stopped after 8.7 s, the increases in temperature and thermal energy enhancement were confirmed through the profiles at 9.0 s.It is noteworthy that the variation of toroidal rotation is interesting in figure 3(d).As the dRsep variation began, the rotation in the pedestal region started decreasing, and after 8.7 s, the rotation in the core region increased sharply.While the increased toroidal rotation gradient may have some effects, we shall not dwell here on this phenomenon in order to simplify among the various phenomena with DN transition.The following analyses in section 3 will have a fixed toroidal rotation profile.

Pedestal and ELMs
The pedestal structure shown in figure 4 is changing in width and height.Each profile was fitted by the tanh form in the pedestal region [72].For CES, which has the higher temporal resolution in the pedestal, profiles of T i was selected just before the ELM crash occurs.Due to the high uncertainty in the edge electron channels, we assumed all the pedestal widths were identical to the T i pedestal and T e = T i in the pedestal.The evolution of the width and height are observed in the pedestal pressure structure shown within figure 4.Although each pressure profile is inaccurate due to the electron channels, the change trend for the electron pedestal during DN transition remained to some extent.In the LSN phase, the pedestal height increased from 7.3 s to 8.0 s while maintaining the edge pressure gradient.After the rapid performance improvement, the pedestal structure changed differently from the general trend.At 8.7 s, the height showed little change, but the width increased significantly, reducing the edge pressure gradient.After the completion of the DN shaping, a high pedestal was formed while keeping the widened pedestal.As seen from the increase in pedestal pressure, the pedestal temperature increased more than the decrease in density during the pedestal enhancement.
Clear transitions in ELM characteristics were observed with the DN shaping.Also, the performance enhancement appeared to be related to changes in ELM characteristics and the dynamics between ELM periods in the higher performance discharge.As seen in figure 2, after 8.4 s, the D α signal became more 'grassy' and showed mixed ELMs. Figure 5 compares ELM bursts in the D α signal and stored energy measured by diamagnetic loops [73].In the LSN phase (see figure 5(a)), regular giant ELMs (red 'x's) that are typical in KSTAR Hmodes were regularly repeated.The ELM frequency remained constant at about 30 Hz, and the stored energy loss by each ELM, ∆W ELM , was maintained at around 15 to 20 kJ.An NBI blip was applied at 8.2 s for measurements, and the stored energy was restored within 0.1 s after the blip.From around 8.4 s, a couple of small ELM bursts (small blue circles) were interposed among type I ELMs in figure 5(b).The small ELM bursts hardly affected the stored energy after the crashes, and multiple small ELMs delayed the next type I ELM crash.Additionally, the type I ELMs followed by small ELMs had moderate ∆W ELM of about 10 kJ.As the pattern repeated until 9.0 s, the stored energy continued to increase.Some small ELMs (big blue circles) had a weak effect on the stored energy, either maintaining it at a steady level or causing a small ∆W ELM of less than 10 kJ.The energy increase stopped at   around 9.0 s, where many small ELM bursts appeared.The energy was saturated with more frequent small ELM bursts.Eventually, the appropriate mixture of the types of ELMs was repeated, saturating the performance enhancement.

Density decrease
The main feature of the performance enhancement discharge is that the density decreased during the DN transition, despite no change in the gas fueling.The density decrease with the dRsep shaping of the enhanced discharge was observed in both the core and the pedestal regions by measurements from TS and multi-chord TCI system in figures 2(b) and 3(b).To analyze the relation between density decrease and performance enhancement, we have compared 35 H-mode discharges controlling the shape from LSN to DN in 2019-2021 KSTAR experiments.These DN transition experiments have been conducted based on the reference scenario for hybrid scenario development with a low density, but some of them still exhibited low performance.The ranges of the plasma current, NBI heating power, and NBI beam voltage were 0.6-0.7 MA, 3.5-4.8MW, and 60-85 kV respectively.All of these discharges did not have strong MHD instabilities that cause performance degradation directly, such as neoclassical tearing modes.The correlation between the performance enhancement and the density decrease is shown in figure 6. Figure 6(a) shows that the DN phases (red dots) are distributed in regions with the lower Greenwald density fraction, n e /n GW , and higher β N in the n e /n GW -β N space, compared to the LSN phase (black dots).The time selected for each points is when the performance remained sufficiently constant in the LSN and DN phases.The sliding window is 0.1 s similar to the KSTAR H-mode confinement time.The error bars in figure 6 are those of discharge 25460.The errors of β N and n e /n GW in the all DN discharges were both within 5%.The relation between the amount of change of the density and β N during the in-shot DN shaping figures out this trend without a tendency on the electron cyclotron power, P EC (see figure 6(b)).

Core MHD activities
A prominent evolution of the core MHD modes was observed during the DN transition.To compare the magnetic fluctuations and the core temperature crashes due to the core MHD activities, the Mirnov coil spectrogram and the electron temperature from ECE at the core channel are shown in figures 7(a) and (b), respectively.In the LSN phase, the sawtooth (ST) instabilities and long-lived modes (LLMs) were irregularly alternated.The LLM is a core MHD instability similar to ST, but it has a long-lived precursor mode, resulting in a weaker and slower crash than ST.While the fishbone (FB) mode was present between each ST and LLM crash, FB became increasingly strong and long-lasting, from around 7.6 s right after the density decrease started.At the end of the DN shaping, at 8.7 s, the FB activity with frequency chirping continued to occur, and the core electron temperature was sustained without any crashes.The intensification of FB implies an increase of the fast ion contents [74,75], consistent with the observation in following section 3.2.
Simultaneously, the ST suppression is related to the change in the core safety factor, q.The suppression of ST could predict an increase in the core q profile.As shown in figure 7(c), the core q obtained from the kinetic EFIT reconstruction was just below unity (or near 1) and increased after 8.7 s.The higher edge q value was due to the increased elongation.Although the result may not be accurate as the equilibria were obtained without the motional Stark effect (MSE) measurement, the relative increase of core q can satisfy the condition for the ST suppression.Periodic minor disruption by ST and LLM crashes can degrade core thermal confinement, as shown in the temperature evolution in the LSN phase.In contrast, the FB mode did not show thermal confinement degradation, ensuring stable performance in the core region.It is suitable for the hybrid  scenario [67], thereby improving not only performance but also advanced operational capability.

Pedestal stability
The enhanced pedestal will contribute to the thermal energy with bulging volume and by acting as a boundary condition of core plasma transport via profile stiffness.To demonstrate these changes in the pedestal structure and ELM characteristics, we have checked the stability in the pedestal region with the PBM theory.The CHEASE code [76] was employed to calculate the numerical equilibria for scanning the maximum value of edge toroidal current density, j ϕ , and normalized edge pressure gradient, α.The bootstrap current was calculated self-consistently using the Sauter model [77] as a constraint while calculating the equilibria with CHEASE.We used the MISHKA-1 code [78] to solve the edge stability for each equilibrium.The stability criterion was determined by γ/ω i, * = 0.25 [79], including the diamagnetic stabilization effect on the pedestal region, where γ is the MHD linear growth rate and ω i, * is the ion diamagnetic frequency at the pedestal region [79,80].
Figure 8 shows the calculated coupled PBM stability boundaries for the LSN and DN phase in the edge j ϕ -α space.The numbers on the figure are the most unstable toroidal mode number, n of the scanned equilibrium.In the LSN phase (see figure 8(a)), the operational point of 7.3 s crossed the vicinity of the stability boundary because we chose the time point just before ELM crashes.Its most unstable n was 3-10, a little closer to the peeling side on the 'nose' region.These results seem similar to some cases generally expected in KSTAR ELMy plasmas [15,67,[81][82][83][84].As shown in figure 8(b), the stability boundary expanded in the high-δ DN phase, consistent with previous studies.In particular, it can be found that the stabilization effect on the current-driven mode is notable so that the upper left side of the boundary, the 'peeling' side, expands significantly.The changes in the pedestal structure during DN shaping were observed at the operational points beyond the LSN boundary, as shown in figure 8(b).However, the operation points in the DN phase (green and red star markers) existed inside the stable region for the stability  boundary of the DN phase (yellow dash-dot line).The most unstable n was about 7, with no significant changes in the peeling mode property.Because we had to use kinetic profiles just before the bigger ELM crash where the T i pedestal measured by CES was clearly observable, the position of operational points could not account for the limitation by small ELM occurrence.Also, the pedestal suppression for the giant ELM crashes implies the interaction between the phenomena in the inter-ELM phase, such as small ELMs and the edge mode described in section 2.3.

Enhancement in the fast ion channel
It appears to be a prerequisite for propagating the impact from the LCFS to the core region, resulting in global enhancement.
If there is no increase in temperature with the decreased density, the thermal pressure should be decreased.However, while the density decreased up to 8.7 s, the shaping maintained or improved the pedestal pressure by increased temperature, which in turn maintained or increased the thermal pressure (see figure 9).As the density level decreases, it may affect fast ion pressure in the core region.Because the NBI is a dominant heating source in these discharges, performance enhancement by the fast ions is essential.The dashed line in figure 9 shows the evolution of the fast ion pressure profile, calculated by NUBEAM [85].In these calculations, we assumed no anomalous fast ion transport.The decrease in density in the core region leads to an increase in the fast ion contents, which may boost the fast ion pressure and total stored energy [67].With the combination of kinetic EFIT and NUBEAM, the contribution of the fast ions to the total stored energy increase is estimated (see table 1).Increasing the fast ion energy and thermal energy simultaneously, the earlier phase with density decrease before 8.7 s shows the higher fraction of the increase in fast ion energy.Regarding fast ion confinement, it remains to be seen whether the fast ion losses depend on the DN shaping.Further dedicated experimentation with more accurate measurements is likely worthwhile.
It is unknown yet under what conditions the dRsep shaping will cause the density decrease.First, one of the most likely causes is the edge mode, which creates the pedestal particle and heat flux generated during the DN shaping.As shown in the spectrogram from beam emission spectroscopy (BES) [86] in figure 10(a), the amplitude of the edge density fluctuation of 35-50 kHz before each ELM crash gradually increased from 7.6 s during the DN shaping.This edge mode was also weakly observed in the magnetic field fluctuation from the Mirnov coil spectrogram (see figure 10(b)).Those, which appear to co-occur with the density decrease, have been consistently observed in other KSTAR experiments, mainly in high-performance hybrid scenarios [84], and showed similar characteristics to the ELM precursor modes observed in JET [87].This mode could increase edge convective transport to reduce the edge density.In terms of the pedestal structure, a wider and higher pedestal mentioned in section 3.1 may have resulted from the transport induced by the edge modes.It would play a similar role to turbulence-driven transport in wide-pedestal QH-modes [88].Further explanations to understand this coherent mode and dedicated experiments are discussed in [84].
Other speculations suggest that the density level will be sensitive to the conditions near the wall and SOL.It is thought that the shaping approaching DN will be greatly influenced by wall recycling and retention near the boundary as a new active X-point occurs on the upper side, especially in a lowdensity regime.In the up-down symmetric KSTAR cryopump design, a particle sink and a new strike point by the upper cryopump can be created.In addition, the 2D transport modeling in the SOL region has suggested several reasons for the density decrease [89].The first is the effect of ion ∇B drift change due to the additional X-point.It induces more convection and lowers the density, changing the convection direction of the ion ∇B drift in the high field side.Second one is related to the E×B drift in the SOL region.The E×B drift increases the recycling rate in the inner divertor.The flow of E×B direction changes according to the magnetic field configuration and Xpoint direction.In the LSN configuration, where the ion ∇B drift direction is toward the X-point, the flow of E×B is inward from the outer divertor near the gas outlet.Meanwhile, when the X-point is located at the opposite side, the flow of E×B is flipped, reducing the recycling rate near the inner divertor.

Integrated modeling
The observed changes related to the DN transition, including those shown in section 3, are simultaneous and interconnected, making it difficult to analyze and quantify the impact of each specific phenomenon separately.To trace the contribution of the thermal confinement enhancement in the core region, we conducted the predictive simulation for the DN transition.In this section, the stationary state of each prediction was estimated self-consistently using the integrated simulation framework, TRIASSIC [90].For the virtual experiments in controlled conditions, the transport solver and external modules within this framework were computed, satisfying consistency internally.ASTRA code [91] was employed to solve 1.5D transport equations, and the CHEASE code was adopted to calculate the magnetic equilibrium.For the neoclassical and anomalous transport model, NCLASS [92] and TGLF [93] codes were employed, respectively.NUBEAM conducted calculations of NBI heating and current drive.The electron cyclotron heating was not used in this discharge.The effective charge, Z eff = 2 was assumed to be radially constant and the toroidal rotation profiles were the same as described above, neglecting the effect of the impurity and rotation here.
To confirm the validity of our model, we checked the prediction both before and after the performance enhancement by the DN transition.Figure 11 shows the predictions for validation.Profiles in the outer region containing the pedestal and the part of the core-edge connection region (known as 'no man's land'), ρ N > 0.8, where conventional transport models generally fail to reproduce, were set as the experimental profiles in figure 3.Then, we assumed our experimental results were stationary states at that moment due to its slow shaping rate.Each simulation representing a specific time point was iterated enough to converge.The temperature profiles were well reproduced compared with the experimental fitted profile, while the core density profile was smaller than the experimental fit profiles.This underestimatation of the density is expected because we did not use particle source models such as neutral gas puffing for the particle transport channel.In the cases shown here, which have a mismatch in density peaking, it is difficult to  predict the exact experiments because the particle source not considered in the calculation can penetrate and affect the core plasma.However, we can confirm the effect of the decrease in density using the density profile calculated here.Although the particle source from the edge is critical during the DN shaping, we presented evaluations without an additional neutral gas model for simplification.The emphasis is on predicting the electron and ion temperature profiles in the core region.Because of the imperfection of our model and the uncertainty of the diagnostics for the density prediction, it is hard to evaluate the particle transport.Therefore, the following analyses solve the transport equation only in the energy channel with the fixed density profile and present the tendency of change in the temperature profile.
For the comparison of the contribution to the performance enhancement, the predictive simulations began with the LSN case with lower energy confinement (7.3 s).We considered the three sets of specific changes from the DN experimental setup and results; (i) changes in the plasma shape itself, (ii) density decrease, and (iii) temperature increase in the pedestal region (see figure 12).In each set of simulations, the other input parameters were fixed to the value of the LSN case: i-1, ii-1, and iii-1 in figure 12.Note that the density profiles given as one of the variables here differ from the experimental profile shown as dashed lines in figure 11.In set ii, We used each density profile obtained in the validation process of TRIASSIC prediction due to the convergence of the TGLF calculation.Without realistic density modeling, we can obtain the tendency of the core energy transport by density decrease.However, we cannot explain the impact of density peaking given in the experimental density profiles.
Figure 13 presents the results of the change in the plasma shape itself.The temperature profile hardly changed with DN transition only.Although the saturation rule of TGLF used here reflects the geometric dependence [94], the boundary shaping only increased the edge q, not propagating to the core region.It indicates that the increase in κ and δ is too small to affect the anomalous transport directly, so that the side effects by the addition of an upper X-point are the main origin of performance enhancement.
The effect of the density decrease is shown in figure 14.In this prediction, the core ion temperature near the magnetic axis, ρ N < 0.4, was increased, while there was little change in the electron channel.It was observed in the CES T i profile in  figure 3(c).This result may be related to an increase in the fast ion contents by density decrease, as shown in section 2.4.As previous studies have suggested [52,67,95,96], the abundant fast ions can stabilize ITG turbulence.A simulation was conducted without considering the main ion dilution effect from TGLF.The temperature profiles in figure 14(b) were predicted using the lower density profile but without the fast ion dilution.As a result, there was no meaningful change in the electron temperature, but the ion temperature decreased to the level at a higher density case (set ii-1 in figure 14).It suggests that the decrease in density reduces core thermal energy transport primarily due to the fast ion dilution.The non-linear electromagnetic fast ion effect can reinforce the stabilization, but our analysis here did not take this into account.
The thermal energy confinement enhancement by the pedestal temperature increase is drawn in figure 15.The observed increase in pedestal temperature was applied as the boundary condition for the core region, keeping the temperature gradient.Notably, the increased core q level by the reduced resistivity is consistent with the experimental result in section 2.5.When the three effects considered here overlapped, the simulation results show the thermal energy transport was either maintained or reduced.Although there seems to be no apparent tendency in the temperature gradient and core energy transport, the performance enhancement by the effect of enhanced pedestal seems to be the most dominant among the three variables we used.

Conclusion
We have investigated the experimental results showing the performance enhancement with the LSN to DN transition using dRsep control in KSTAR.DN transition can affect more than just the addition of the upper X-point.It is likely to cause side effects that enhance performance.Also, in this process, various phenomena occur complexly as the edge boundary geometry changes and the impact propagates to the core region.The experimental results showed that the changes in the pedestal structure, ELM characteristics, decreased density, and core MHD activities affect the performance enhancement.It was observed that a wider and higher pedestal and more grassy and mixed ELM characteristics.The decreased density was a major characteristic associated with performance enhancement.The accompanying transition in core MHD activities was identified from ST to FB instabilities.
Edge stability analysis showed that the strong shaping expands the stability boundary, but it still left the question about the onset of small ELMs.Also, the density decrease influenced the fast ion content in the core region.The separation of effects of shaping, density decrease, and pedestal temperature increase on the thermal energy confinement was conducted by predictive integrated simulations.In the simulations, the DN shaping itself did not affect the core thermal energy transport, only the edge q value.The major impacts of the density decrease and the pedestal temperature increase were found.The increased fast ion content by the density reduced the ion transport in the core region by ITG stabilization due to the dilution effect.The temperature increase in the pedestal channel as a boundary condition of profile stiffness had more dominant contributions than others.
It also influenced the core q profile shape, similar to the experimental observation.While the changes overlap, it is worth noting that performance did not sharply improve at the moment dRsep becomes zero but rather slowly rises with the shaping control process, even after around 0.3 s of the DN shaping completion.There might be a possibility for gradual enhancement if the pedestal improvement, which is confirmed to be responsible according to the integrated modeling, occurs slowly as the ELM cycle changes during the transition.The conditions of its onset have not yet been known.Further studies about ELM dynamics are required to understand it further.This slow enhancement should be considered essential in devices with longer pulse duration, such as superconducting tokamaks.
So far, we have addressed major findings for the origin of performance enhancement, neglecting several points, such as the change in the rotation and impurity profiles.For example, the effect of the E×B shear rate is expected to contribute to both the core and pedestal region [67,97], along with the variation in the toroidal rotation profile observed in our DN shaping discharges.Also, changes in the wall interaction by the DN shaping may adjust the impurity concentration inside the plasma, and thus changes in the ion density profile may occur.Another issue is the cause of the density decrease, which has yet to be thoroughly identified, but edge mode occurrence and SOL transport/wall condition change were speculated.
The mechanism of how weak ELM bursts occur in the DN phase and how they can affect performance was also unknown.When a type I ELM is expected to occur, replacing it with this weak ELM seems appropriate for reducing ∆W ELM , but too frequent small ELMs might slow down the energy increase.Perhaps, the manipulation by these complex ELM dynamics is one of the necessities for performance enhancement.Unfortunately, there was a lack of faster profile diagnosis (below 1 ms) to capture the pedestal changes with the mixed ELMs.Still, we cannot yet present a more detailed picture describing these ELM characteristic transitions.It is required to further analysis, such as nonlinear pedestal evolution simulation [98], for the KSTAR grassy ELMs.
DN shaping can be applied to easily optimize improving performance from the plasma edge region under conditions with wider and higher pedestal formation and decreased density.It is essential to enhance the plasma using the propagation B. Kim et al from the edge to the core region while taking into account the synergetic effect, thus core-edge interplay, which was not addressed in this paper.Together with distributing the large power by DN shaping, this approach can be utilized in an efficient operational scenario for future fusion reactors having the DN divertor design [99].

The 1 -
D kinetic profiles are compared during the four stages in the DN transition to grasp the changes in the process in detail and estimate the increase in thermal energy.Each time point at 7.3 s (LSN, dRsep = −1.4cm), 8.0 s (LSN, dRsep = −1 cm), 8.7 s (DN, β N = 2.4), and 9.0 s (DN, β N = 2.7), was compared with 1-D measurements in figure 3. The electron temperature, T e , and the electron density, n e , taken

Figure 2 .
Figure 2.An example of a discharge presenting performance enhancement due to the transition from the LSN to the DN configuration in KSTAR; Shot 25460.Time trace of the main parameters during the shaping evolution; dRsep, q 95 , β N , H 89 , and 2l i , toroidal Dα signal, line-averaged electron density ne, electron temperature Te at the centre, and neutron emission rate.

Figure 4 .
Figure 4. Edge ion temperature and the pressure profile in the pedestal region, ρtor > 0.8.

B. Kim et al Figure 5 .
Figure 5. ELM characteristics from Dα signal and stored energy from diamagnetic flux measurement during the shaping evolution (shot 25460).

Figure 6 .
Figure 6.Comparison of performance enhancement and density drop observed in conducted DN experiments.(a) β N versus Greenwald density fraction, ne/n GW for stationary LSN and DN phases.(b) ∆β N vs. ∆(ne/n GW ) after the DN transition.Each point corresponds to one discharge.

Figure 7 .
Figure 7. Core MHD activities from (a) the spectrogram from the Mirnov coil signal and (b) the evolution of core temperature from ECE measurement near the magnetic axis.(c) Safety factor profiles reconstructed by the kinetic EFIT including MSE measurement.

Figure 8 .
Figure 8. PBM growth rate (color map) and stability boundaries with equilibrium of (a) the LSN and (b) the DN phase of shot 25460.The white line on the contours is for the LSN shape (dashed) with the stability criterion, γ/ω i, * = 0.25.The yellow line is for the DN shape (dash-dot).The star markers are the operational points of 7.3 s (black in (a)), 8.0 s, 8.7 s, and 9.0 s (blue, green, and red in (b)).

Figure 9 .
Figure 9. Thermal pressure, P th and fast ion pressure, P fast profiles during the DN transition; the thermal pressure, P th in solid and the fast ion pressure, P fast in dashed lines.The pressure profile in the pedestal, ρtor > 0.8 is plotted within.

Figure 10 .
Figure 10.Edge modes observed in the beam emission spectroscopy (BES) (a), and the Mirnov coil (b) in shot 25460.

Figure 11 .
Figure 11.Prediction of the electron and ion temperature and the electron density profiles in 7.3 s (LSN, before performance enhancement) and 9.0 s (DN, after performance enhancement) cases with performance enhancement.Their experimental fits are shown in dashed lines.

Figure 12 .
Figure 12.Three input variable sets for the simulations.(a) Plasma shape, (b) density, and (c) pedestal temperature profiles.

Figure 13 .
Figure 13.Prediction results with DN shaping variation only (set i).Electron and ion temperature, and safety factor profiles.

Figure 14 .
Figure 14.Prediction results from the density decrease (set ii).(a) Electron and ion temperature, and safety factor profiles.(b) The lower density case (set ii-3) is recalculated, excluding the fast ion dilution effect from the TGLF model (magenta lines).

Figure 15 .
Figure 15.Prediction results from the pedestal enhancement (set iii).Electron and ion temperature, and safety factor profiles.

Table 1 .
Total stored energy, W total of each time in figure9, the increase of stored energies (thermal energy W th and fast ion energy W fast , where W total = W th + W fast ), and the fraction of the fast ion energy contribution.