Understanding JET-C quiescent phases with edge harmonic magnetohydrodynamic activity and comparison with behaviour under ITER-like wall conditioning

An analysis of edge localised mode-free (quiescent) H-mode discharges exhibiting edge harmonic magnetoydrodynamic activity in the JET-carbon wall machine is presented. It is observed that the otherwise quiescent pulses with multiple-n harmonic oscillations are sustained until a threshold in pedestal electron density and collisionality is crossed. The macroscopic pedestal parameters associated with the quiescent phase are compared with those of a database of JET-ELMy discharges with both carbon and ITER-like wall (ILW). This comparison provides the identification of the existence regions in the relevant pedestal and global plasma parameters for edge harmonic oscillations (EHOs) in JET plasmas. Although the ELMy database scans pedestal collisionality and β values typical of ET-carbon quiescent operation, shaping and current are not simultaneously compatible with EHO existence. Nevertheless, ILW operation with JET-carbon quiescent-like parameters could in principle be achieved, and improved pedestal performance could be observed in more recent JET-ILW pulses.


Introduction
The most promising scenarios for achieving efficient controlled thermonuclear fusion in tokamak machines are the so called high-confinement (H-mode) regimes. Such scenarios show long energy confinement times and are typically characterised by the presence of sharp and narrow plasma edge pedestals, both in mass density and temperature. Unfortunately, the associated strong radial gradients favour the appearance of short wavelength magnetohydrodynamic (MHD) perturbations called edge localised modes (ELMs) [1]. These sudden and violent events are associated with rapid energy and particle expulsions which deposit intolerable heat loads on plasma facing components. In addition to severe plasma contamination, ELMs can significantly reduce machine lifetime. Therefore, it is of crucial interest to attain high-performance scenarios without the deleterious presence of ELMs [1].
One of the most promising intrinsically ELM-free regimes is the so called quiescent H-mode. In this regime, which shares with the standard H-mode large edge pressure gradients and long energy confinement times, ELMs are avoided and replaced by continuous low-n mild MHD perturbations, and the associated peak energy and heat loads on the plasma facing materials are significantly lower compared to ELMy regimes. These edge harmonic oscillations (EHOs) are well localised within the edge region of large gradients (pedestal) and feature multiple n toroidal harmonics with a rather long lifetime, of the order of 1 s [2]. EHOs have been observed in DIII-D, ASDEX-U, JT60 [3][4][5][6], and JET [7]. In [7] such oscillations are called Outer Modes. Since the MHD dynamics described in [3,4,7]. have the same characteristics, for the sake of clarity, hereafter we will refer to such oscillations as EHOs 13 . As observed in [7], EHOs have several common features with the low-n type-I ELM precursors studied in [10].
It is observed that EHOs in JET are prone to develop in the early phase of the discharge, during the density ramp when ion and electron temperatures reach their highest values at the pedestal top. It is also observed that above a critical value of the pedestal density (∼5 × 10 19 m −3 ) the quiescent phase with EHOs abruptly ends and ELMs appear. We point out that quiescent regimes in DIII-D, ASDEX-U and JT60-U experiments [3][4][5][6] usually operate with lower values of the pedestal top density compared with JET, typically three or four 13 Historically, Outer Modes refer to low-n (mainly n = 1) MHD oscillations with frequencies ∼10 kHz observed typically within the outer 20% of the plasma [8,9] (broader compared to EHOs). In [9], Outer Modes refer to current driven external kinks. times smaller. Although for some discharges EHOs can reemerge after an ELM crash, the quiescent phase is not usually recovered after the first ELM is triggered.
Thus, the aim of this paper is to detail and characterise the existence conditions for EHOs in JET, in order to determine the pedestal features required to guarantee the accessibility to quiescent regimes with EHOs, and potentially in machines with a metallic wall. These existence conditions are assessed by comparing global and pedestal parameters, namely electron β, collisionality, q 95 and triangularity, of quiescent with EHOs (Q-EHO) shots with the ones extracted from a carbon (C) and ITER-like wall (ILW) EUROfusion pedestal database of ELMy discharges [11,12]. It is found that the operational regime of Q-EHO plasmas identified by the parameters given above does not overlap the one explored by the ELMy database. This holds in particular for ILW plasmas where the electron temperature, which plays a crucial role, is significantly smaller compared to carbon wall discharges. We finally point out that some indications of brief edge coherent activity have been observed in recent hybrid JET-ILW shots with high pedestal performance. Although these oscillations have been observed transiently, efforts are now focussed on steadily sustaining this behaviour in metal machines [13].
Thus, the paper is organised as follows: In section 2 we describe the JET-C experimental set-up and the typical features of the EHOs which are observed during the quiescent phase. In section 3 we analyse a database of JET discharges with a carbon and ILW, which although exhibiting quiescent compatible pedestal parameters are in ELMy regime. This is done by inspecting the associated physical global and pedestal parameters (e.g. collisionality, temperature, etc). These parameters are then compared with the ones observed in the Q-EHO JET-C pulses analysed in section 2. By assessing the differences between JET-C plasmas with EHOs and the ELMy database, we infer the conditions that have to be met to achieve quiescent operation, and in particular whether such conditions are compatible with a metallic wall. Finally, a discussion of the results and concluding remarks are given in section 4.

Experiment and analysis setup
We analyse a set of four JET-C discharges showing a quiescent ELM-free phase in which coherent edge EHOs are observed. These pulses feature a hot-ion H-mode phase [7,14], obtained by reducing particle fuelling from external sources and wall with an initial operation at low plasma density and fairly high co-current NBI power yielding very high ion temperatures. This was specifically designed for achieving high pedestal temperatures. Bolometry and D α emission confirm the ELMfree phase, and the EHO identification is achieved by matching the mode rotation frequency inferred from magnetic signals with the one obtained by edge electron temperature fluctuation measurements [15] (see the next subsection for further details). This provides a unique mode identification for toroidal spectral structure and spatial localisation. The requirements for the pulse selection are (i) a relatively long quiescent phase duration, and (ii) diagnostic signal clarity, i.e. all of the EHO footprints must be clearly visible on all the relevant diagnostics. The pulse numbers of the four discharge analysed in this work are listed in table 1, which also reports the vacuum toroidal field B t at the plasma geometric centre, the plasma current I p , β N (see [3] for the definition), q 95 , elongation κ and triangularity δ. The NBI input power is also given, as well as the time window t 1 − t 2 within which the analysis is performed. The EHO appears approximately at t 1 , whereas t ELM indicates the time when the quiescent phase with EHOs is lost and the plasma enters the ELMy regime. At time t 2 the plasma is deep into the ELMy phase. Other quiescent discharges sharing very similar features with the ones reported in table 1 with a likely EHO activity, have not been analysed because of the lack of clarity on some of the diagnostics signals.
Error field correction coils (EFCCs) for ELM control were employed in discharges #78012 and #78014. In both pulses the coil current was ramped up and down in 100 ms, with the current flat-top of ±0.5 and ±1 kA for shots #78012 and #78014 respectively lasting from t = 14.1 s to t = 15.9 s (the plus/minus sign refers to 1-5/3-7 octants). No EFCCs were employed in pulses #75411 and #79455. The ICRH heating is switched on only at the very end of each pulse, so that we regard these shots as Ohmically and NBI heated.
All pulses in table 1 exhibit similar behaviour, in which the electron temperature increases all across the plasma core region during the early phase when the NBI is switched on, and whose power is steadily maintained for the whole discharge (cf figure 8 in section 2.3). At the same time, the plasma density n e is ramped up, reaching the stationary core value  width, are detailed in [16]. The density rise is accompanied by an increase of the effective charge Z eff which ranges approximately from 1.8 to 2 during the quiescent phase, whereas during the ELMing stage reaches values up to 2.5. We observe that in pulse #78014, the EHO disappears at t = 14.23 s despite the low pedestal density, and a sudden increase of n ped e occurs after the appearance of the first ELM. The EHO loss may be caused by variations in the global temperature profile which are not captured by the local pedestal analysis (see section 2.5).
As the mass density is ramped up, the toroidal rotation v tor (obtained from charge exchange measurements, cf figure 3) is observed to reduce whereas β N remains approximately constant. Note that in JET the NBI is co-current [17]. In the time window indicated in table 1, ion and electron temperatures tend to decrease, with T i exhibiting smaller gradients in the pedestal compared with T e . T i and v tor are found to have similar radial dependencies. We point out that during the quiescent phase the ion temperature at the pedestal top is slightly larger on average compared with T e , the former taking values Evolution of the pedestal ion (red) and electron (blue) temperatures for the discharges of table 1. T i is obtained from charge exchange measurements, whereas Te from HRTS. Note that the ion temperature has a more pronounced decrease during the pulse evolution compared to Te. of about 2 keV whereas the latter is ∼1.4-1.5 keV at the EHO onset. The two temperatures get closer after the appearance of the first ELM.

Characterisation of MHD dynamics
Very rich MHD dynamical behaviour is observed during the early NBI heating phase as clearly shown in figure 4. EHOs however behave similarly in all four shots of table 1 with a lifetime of ∼0.5-1 s. In pulse #78014, after an internal (core) temperature crash at t = 13.68 s, a steady EHO is sustained for ∼400 ms starting at t = 13.8 s (cf table 1). Each toroidal harmonic with mode number n of the EHO rotates with frequency nΩ ped where Ω ped is the rotation frequency at the pedestal top [7]. After the steady EHO phase is lost, short lived pre-ELM EHO bursts with multiple low-n harmonics appear from 14.3 to 14.5 s. We point out that similarities between EHOs (or Outer Modes) and the low-n ELM precursors studied in [10] have been pointed out in [7]. The latter appear as multiple n coherent oscillations localised at the plasma boundary, and were observed frequently in hot-ion H-mode regimes [10]. The poloidal spectral structure of these low-n ELM precursors features poloidal mode numbers comparable or slightly larger than q 95 , and we expect EHOs to have similar characteristics. Indeed, as described in [7], the magnetic signal was found to have a strong m = 4 component. Note that this is also consistent with recent findings in [18], where a thorough description of the magnetic spectrum is given.
During the density ramp-up, when the electron temperature is the highest, a clear signature of multiple n harmonics all equally spaced in the frequency domain appear on the magnetic diagnostics. In order to assess the radial location of these MHD modes, and therefore identify their EHO-like nature, in analogy with [5,15] the magnetic measurements are compared with the electron cyclotron emission (ECE) signals. The EHO is uniquely identified by matching the rotation frequencies of the various harmonics measured by the two diagnostics. The ECE channel distribution and location in the major radius of the outer midplane is given in figure 5. Note that for all the shots listed above, channels 54 and 66 are usually associated with pedestal measurements (odd number channels were not available). Note that in all discharges listed in table 1, the pedestal ECE emission is, with a good approximation, associated with channel 66 (highlighted in red). Channel 56 is the nearest-non pedestal channel, so that it is always inspected to check the edge mode localisation. It is important to point out that even channels only are associated with the fast ECE data acquisition. The separatrix position is denoted by the dashed vertical line. A typical spectrogram of ECE near-pedestal signals (channel 66 of figure 5) is shown in figure 6. The electron temperature fluctuation with the many-n harmonic structure displayed in figure 4 is clearly recognisable. It is found that the ECE signal does not propagate further beyond the pedestal shoulder. Indeed, as clearly shown in figure 7, the ECE trace of the EHO is not visible inside channel 56 which is the first available channel in the core region outside the pedestal. DIII-D and JT60-U experiments [6,19] also reported similar behaviour, which resembles the dynamics of low-n ELM precursors studied in [10]. These fluctuations have been shown to be localised within the pedestal radii and not to extend radially beyond the pedestal shoulder, giving us confidence on the pedestal radial localisation of the EHO in JET, which must indeed be extremely narrow.
It is worth noting that EHOs in DIII-D, ASDEX-U and JT60 have been observed in low pedestal density plasmas of the order of 1 × 10 19 m −3 and currents of 1 MA [3][4][5][6], whereas JET quiescent plasmas can be sustained up to fairly large values of the pedestal density (n ped e,crit ∼ 5 × 10 19 m −3 , cf figure 1). Although JET operates at higher plasma current, this behaviour is not fully explained as local quantities such as collisionality, the electron β e , Z eff seem to be comparable across these machines [4,20].
Furthermore, we notice that EHOs are not affected by external magnetic perturbations, at least up to 1 kA of EFCC current. Indeed, the EHO dynamics observed in discharges #75411 and #79455 with no active EFCCs, is essentially equivalent to the one of pulses #78012 and #78014 in which EFCCs were applied. Also, in shots #78012 and #78014, EHOs occur well before EFCCs are applied at ∼14 s, and disappear well before the EFCC current is switched off at ∼t 2 . This suggests that the EHO must be driven primarily by internal, i.e. within the last closed flux surface, mechanisms.
It is worth pointing out that the ELM appearance does not necessarily imply that the quiescent regime cannot be recovered. Indeed, in discharge #78012 a first quiescent phase lasts until t ≈ 14.3 s when an ELM occurs and the EHO is lost; after this ELM event, the EHO appears again at t ≈ 14.45 s and lasts until t ELM . In pulse #79455 instead, the EHO disappears spontaneously at t ≈ 14.64 s, whereas the quiescent state persists until t = t ELM .

The role of plasma rotation
We now argue that the toroidal rotation affects weakly the appearance of EHOs. This is because we observe that the toroidal rotation frequency at the pedestal top drops after the appearance of the first ELM as shown in figure 8, which also indicates that EHOs exist within a wide range of pedestal rotation frequencies. Moreover, although a connection between toroidal (carbon) rotation shear and Q-EHO phase was established in [7], DIII-D results of [21] show that the accessibility of the quiescent phase is almost independent of the toroidal (carbon) rotation shear. It is also worth stressing that there could be a consistent difference between carbon and mainion species rotation profiles in the pedestal region, the latter exhibiting significantly weaker gradients [22][23][24]. As such, the rotation of the carbon impurity may not be a good proxy for inferring the main ion rotation properties. Furthermore, nonlinear MHD simulations with the JOREK code [25] found that toroidal flows have a weak effect on the destabilisation and saturation of modes which might be related to a Q-EHO phase. Interestingly, we notice that similar edge localised oscillations with a dominant n = 1 component have been recently reported in Alcator C-Mod in low-collisionality and high pressure pedestal regimes [26]. No NBI was employed in the C-Mod experiments [26], supporting our claim that these edge fluctuations do not depend explicitly on plasma toroidal rotation. In conclusion, the experimental evidence in JET, also supported by the results presented in [6,21] and numerical modelling, gives us confidence that other physical effects might be more relevant for the EHO appearance.

Radial electric field during the quiescent and ELMy phases
As pointed out in [21], one of the key parameters which determine the accessibility to the quiescent phase is the edge E × B flow shear. This flow manifests itself as a plasma rotation mainly in the poloidal direction, and its strength is proportional to the radial electric field. From the radial ion force balance equation [27], allowing for plasma shaping through elongation, we have near the edge where p i = n i T i with n i the ion density, e the ion electric charge, κ the plasma elongation, a and R 0 the minor and major radii respectively, Ω tor = v tor /R the toroidal angular frequency, v pol the poloidal velocity and r a flux label (for the geometry of the beam injection and sign conventions we refer to [17,28]). Writing R = R 0 + r cos θ, we have d/dr = d/R at θ = 0. This allows to simplify the analysis by taking all the relevant quantities as a function of R on the equatorial outer mid-plane. Figure 9 shows E r averaged over the quiescent and ELMy time windows as a function of the major radius. The E r well takes values comparable to those observed in DIII-D [19,29] of the order of ∼100 kV m −1 , although no strong variations of E r are observed when transitioning to the ELMy phase [30]. This finding suggests that in the four discharges of table 1, the pedestal density evolution should be the main actor responsible for transition from quiescent to ELMy regime. This, indeed, is discussed in the next subsection.

Pedestal density and temperature conditions for the Q-EHO phase accessibility
Thus, to further investigate what conditions favour the EHOs existence, we study the localisation of the discharges of table 1 in n ped e − T ped e space. Figure 10 shows the instantaneous pedestal values of density and electron temperature during the time window t 1 − t 2 for the four shots considered. We observe that the Q-EHO phase exists in the region of low-density with temperatures ≳1 keV. Note that most of the high pressure values associated with an ELMy pedestal belong to pulse #78014, and these may be connected with the EHO-like bursts prior to ELMs shown in figure 4. During the discharge, the plasma evolves towards an ELMy regime which loses the EHOs. Here ν e * follows the definition of [32] ν e * = 6.921 × 10 −18 q 95 R 0 n e Z eff ln Λ e T 2 e ϵ 3/2 , where ε is an effective inverse aspect ratio defined by ϵ = a R0 √ 1+κ 2 2 with R 0 and a the major and minor radii respectively with a/R 0 ≈ 0.33, and ln Λ e is the Coulomb logarithm. As pointed out in [5], plasmas with a higher percentage of impurities may require smaller pedestal densities in order to maintain the pedestal electron collisionality sufficiently small. Notice that the increase of pedestal collisionality is likely to yield a reduction of the bootstrap current, and a consequent increase of the local magnetic shear [32][33][34][35]. Hence, the collisionality threshold may be associated with a critical value of the magnetic shear below which EHOs can develop [36], in accordance with recent analytic and numerical modelling [18,[37][38][39][40].
Thus, EHO dynamics in JET appear to be governed primarily by β ped e and ν ped e * , both identifying a threshold for the EHO appearance, i.e. β ped e > 0.2% and ν ped e * < 0.3 (see figure 11). Other parameters, such as q 95 and the pedestal values of the magnetic shear, may play a role in the EHO triggering and their effect on the edge MHD behaviour this will be discussed in the following section.

JET-C and ILW ELMy database pedestal analysis
The aim of this section is to analyse the pedestal features of a database of JET discharges with carbon and ILW that, although partially fulfilling the parameter requirements for Q-EHO operation, are in ELMy regime. This would help in identifying if further hidden parameters, other than ν ped e * and β ped e discussed in the previous section, determine the accessibility to the ELMs/no-ELMs phase, and assessing if Q-EHO scenarios could potentially be reproduced in metallic machines.
The JET-C/ILW database under consideration consists of 1216 shots in high performance ELMing H-mode, divided into 360 shots with carbon wall and 856 shots with ILW [12]. Further details about this database can be found in [12]. The pedestal values of the associated physical quantities are averaged over a time window of 1-2 s during an inter-ELM stationary phase. Contrarily to the analysis of the previous section which focussed on the discharge evolution in the early phase when density is ramped (cf figure 1), the time window of all shots in the ELMy database is taken during the steady state flat-top. It is worth stressing that if EHOs exist in a transient phase, i.e. they are robust, they should be expected to be also observed In analogy with figure 10, the pedestal electron temperatures plotted versus density for the shots in the database are shown in figure 12. At first glance, ILW discharges tend to occupy the lower part of the n ped e − T ped e plane having on average lower temperatures, while exploring regions of similar densities. We observe that for these pulses the ion and electron temperatures have comparable values (i.e. T i ∼ T e ). We point out that within this database, high electron pressure regimes above 12 kPa are not explored by ILW discharges. As a direct consequence, from an inspection of the associated pedestal electron collisionality, we clearly see that ILW discharges tend to have higher ν ped e * values. Nevertheless, a reasonable number of ILW shots lie in a region of low to moderate collisionality at fairly high (low) electron temperature (density), where EHOs might be expected to exist.
As a first check, it is instructive to see whether the JET-C discharges in the database lie in the same region of the n ped e − T ped e parameter space of the quiescent ones studied in the previous section. Hence, the n ped e − T ped e data points of This is shown in figure 13. It is immediate to notice that quiescent discharges occupy the region of the parameter space of high temperature and low density with the electron pressure above ∼7 kPa at low collisionality (ν ped e * ≲ 0.3 in line with DIII-D results [4]). It should also be noted that in figure 13 an upper limit in pressure ∼14 kPa appears where no EHOs are observed. As pointed out in section 2.5 (cf figure 11), beyond the small collisionality requirement, these results may suggest that quiescent EHOs phases are accessed when a threshold in the pedestal pressure is crossed. In addition, by comparing figures 12 and 13, the n ped − T ped e parameter space explored by JET-ILW shots seems not to overlap with the one of the JET-C Q-EHO discharges.
However, we notice that JET-C ELMy plasmas are found in regions where EHOs are expected to exist, i.e. despite having similar pedestal characteristics with the quiescent ones. Thus, we argue that other parameters must play a key role in determining whether or not the quiescent phase is accessible. We point out that when interpreting EHOs as pedestal localised pressure driven MHD instabilities [37,39,41], similarly to ballooning modes and to some extent Mercier modes, it is more appropriate to compare the pedestal β value rather than the pressure alone. This indeed is shown in figure 14, where the pedestal values of β e , defined as in section 2.5, of the carbon and ILW database discharges are plotted against of ν ped e * (cf figure 11(c)). The green points in figure 14   It is immediate to notice that JET-C discharges have a favourite access to high β ped e and low collisionality (ν ped e * ≲ 0.3) regions, whereas ILW shots tend to cluster mainly at high collisionality with lower β ped e (cf figure 12). In the space identified by the parameters given above, namely collisionality and β, the set of four discharges given in section 2 occupies the high-p ped e low-ν ped e * region which is well covered by the JET-C ELMy database, contrarily to the ILW shots which barely border it. Hence, as mentioned previously, other parameters must play a role in the determination of the quiescent/ELMy state.
Let us first point out that apart from the pedestal height (i.e. n ped e , T ped e and p ped e ) and collisionality requirements, experiments in DIII-D, ASDEX-U and JT60-U showed that EHOs are obtained in plasmas with a sufficiently large edge-wall clearance. In DIII-D, ASDEX-U and JET, the plasma-wall gap is enhanced by having large values of triangularity, both upper and lower, whereas JT60 plasmas were able to operate with quiescent phases at low δ as long as a sufficient plasma-wall clearance was provided [6]. The necessity of the plasma-wall detachment is found to be a key ingredient for the EHO excitation in analytic modelling [37,39] which also predicts the dependence upon the value of q in the pedestal in accordance with numerical simulations of JET-like plasmas [41,42]. This dependence manifests itself by favouring the appearance of EHOs when q has a weakened shear and is close to a low m/n rational, as shown by the exfernal model of [37][38][39]. We might therefore expect EHOs when q 95 , or more precisely q in the pedestal, takes particular values.
Thus, we identify four parameters that are likely to determine the accessibility to the quiescent regime with EHOs: β ped e , ν ped e * , δ and q 95 .
Note that apart from triangularity, all these quantities are local in nature, i.e. they specifically identify the pedestal conditions. It is worth to point out that although the plasma-wall clearance could be affected by elongation, the value of κ associated with the discharges in the JET-C/ILW database does not significantly differ from the one of the pulses reported table 1. As such, elongation can be regarded as a fixed parameter. We expect that EHOs can be excited if the quantities associated with the database of figure 12 take simultaneously the same values of, or are close enough to, the ones of the quiescent discharges of table 1. Hence, for each discharge of the ELMy database we plot the parameters listed above against each other, analysing how they distribute with respect to the ones of the quiescent pulses of section 2. Since the β ped e − ν ped e * space has already been analysed in figure 14, we plot in figures 15-17 the remaining combinations of the parameters given in (1).
Let us point out that the ELMy pulses in the database are only a small subset of all the JET discharges within the range indicated in figure 12. Hence, the appearance of clusters in figures 15-17 is not due to the lack of experimental points, but rather to the database selection process.
Analysing where Q-EHO shots from table 1 locate in the parameter space identified by combining pair by pair the physical variables given in (1), shows (cf figures [14][15][16][17] that there is some overlapping with carbon ELMy pulses, whereas such overlap is less pronounced for the ILW ones. Such overlapping is further reduced when considering database discharges with ν ped e * < 0.3 and β ped e > 0.2%, and this is even more evident for ELMy ILW shots as shown in figures 15(b)-17.
In particular, we notice from figures 16 and 17 that ELMy database discharges with EHO compatible β ped e , ν ped e * and δ but different values of q 95 (cf table 1) do not exhibit quiescent phases. Thus, we argue that the dependence upon q has to be a key ingredient in the mode excitation [31] (cf section 2). The role of q 95 , however, is not clear since experiments in DIII-D have been observed not to have preferred values for q 95 ,  whereas ASDEX-U and JT60-U achieved quiescent phases only in a narrow range of q 95 [31] similarly to the JET findings shown in this work. Nevertheless, as previously discussed, if EHOs are to be considered as a pedestal MHD phenomenon, qpedestal rather than q 95 is the relevant parameter which should be used for determining the Q-EHO phase access, among with the local value of the magnetic shear [40]. Note that q-pedestal and q 95 have been shown to differ appreciably [43]. However, extracting information on the pedestal values of q and its shear proves to be rather difficult, and further analyses are required to characterise better the experimental role of the two parameters.
Hence, from the results presented above and inspecting particularly figure 14, we are induced to infer that EHOs have not been observed in the ELMy database of figure 12 because none of the pedestal parameters had simultaneously the values that are specific of the JET-C quiescent shots of table 1.
Finally, although EHOs have not been observed in the aforementioned database of ILW discharges, there are some indications from more recent experiments that EHOs might be seen with a metal wall. Indeed, brief edge magnetic oscillations have been observed in high performance pulses with strong NBI heating. The associated macroscopic parameters, e.g. shaping and pedestal values, are similar to the ones of the quiescent JET-C discharges of table 1. Further analyses are however required to have a unique mode identification. Additional support to the possibility of EHO activity in metallic machines comes from novel ASDEX-U findings, where EHOs have been observed, although transiently, in co-NBI heated pulses with the NBI applied before entering the high confinement phase [44]. Moreover, more recent experimental observations in ASDEX-U reported the presence of EHOs lasting for several hundreds of milliseconds, suggesting the possibility that such oscillations could be steadily sustained [13].

Conclusions
In this work we presented a detailed analysis of quiescent JET discharges exhibiting EHO-type MHD activity. These edge oscillations have been observed in the early phase of the discharge, during the density ramp when strong NBI heating is applied to a low density plasma. In line with previous results on other devices such as DIII-D and ASDEX-U, EHOs are observed at low collisionality [31] with the pedestal characterised by high temperatures and low density. The pedestal densities achieved in JET are three or four times larger compared to the ones required to sustain EHOs in DIII-D, ASDEX-U and JT60-U, despite the fact that similar values of collisionality, β e and Z eff are reported across these machines. The density ramp is accompanied by a decrease of the pedestal temperature, with the knee of T ped e exhibiting an inwards shift leading to an increase of the pedestal collisionality. Plasma toroidal rotation does not seem to have a significant role in determining the EHO stability (cf figure 8), and this also holds true for external magnetic perturbations generated by EFCCs with currents up to 1 kA. This leads us to infer that the EHO dynamics should be primarily determined by internal mechanisms driven by the pedestal values of temperature, density and current, the latter being strongly affected by bootstrap contributions whose fraction is in turn dictated by plasma temperature and density.
The quiescent phase with EHOs, which is sustained until ν ped e * ∼ 0.3 and with β ped e > 0.2%, is lost when pedestal top plasma density reaches a threshold value of ∼4−5 × 10 19 (m −3 ). No EHOs are observed below a pedestal electron pressure of ∼7 kPa. The pedestal β value recorded in JET is in line with the one achieved by DIII-D [4], i.e. ∼1% by allowing for T i being approximately 30% larger than T e . The pedestal features associated with JET-C Q-EHO discharges have been consequently compared with the ones of a database of ELMy pulses with both carbon and ILW conditioning. Database pulses with ILW tend to operate at lower temperature and pressure compared to the carbon ones, and this is reflected by having a larger fraction of ILW shots at high collisionality. As a direct consequence, Q-EHO pulses overlap regions in the n ped e − T ped e space that, within the ELMy database, are explored primarily by carbon and not ILW pulses. Beyond β ped e and ν ped e * , other important parameters which can influence the accessibility to the quiescent phase are the plasma-wall detachment, determined primarily by the triangularity, and the plasma current, i.e. q 95 .
It is found that database discharges with quiescent compatible ν ped e * and β ped e , fail to have triangularity and q 95 values falling in the range of those observed in JET-C with EHO pulses. On the other hand, ELMy pulses with δ and q 95 values close to the ones of quiescent scenarios do not have the required pedestal collisionality and pressure to sustain the EHO. Although the role of q 95 is not clear, as other devices achieved quiescent operation with EHOs without a preferred value of q 95 , we are led to infer that the pulses of the ELMy database did not exhibit EHO activity because ν ped e * , β ped e , δ and q 95 did not have simultaneously the specific values characterising QH discharges observed in JET-C. Nonetheless, by inspecting the ELMy database, we see that ILW discharges have the potential to explore parameter regions which are compatible with quiescent plasma operation, and indications of transient edge oscillations have been observed in more recent JET-ILW pulses. This is also supported by findings in ASDEX-U [13] which seem to point clearly to the possibility that QH regimes with EHOs could be accessible in metallic machines.
Further analyses are nevertheless required to better understand the impact of the pedestal q and other parameters, such as wall conditioning and input power, on the MHD dynamics in these scenarios, aiming at the steady sustainability of quiescent regimes in metallic devices.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.