Above a certain ion heat flux [29] a transport barrier forms at the plasma edge (ETB) and H-mode is accessed. The factor two improvement with respect to L-mode makes H-mode the preferred operation scenario for future tokamaks. Despite being known since 1989 [30], key aspects of the basic physics understanding are still missing and no predictive theory exists, though it is widely accepted that turbulence suppression by sheared E × B flows is a key ingredient and is ultimately what sustains the H-mode.

Though understanding of turbulent transport in tokamaks has improved greatly over the years, we are still far from predictive first principle modelling of future devices, in particular for STs, and rely on empirical scalings of the energy confinement time τ_E. The data at low aspect ratio and high β will help to improve the scalings by expanding the parameter space. For future STs the scaling of τ_E in H-mode with collisionality is the most important, as this requires the furthest extrapolation. This is one reason to increase I_p and B_tor as well as the heating in MAST-U (see Section 12.3).

MAST has a large tank volume. The neutral gas buffer in this volume dominates the fuelling, in particular fuelling by LFS gas puff. For such plasmas the particle confinement on MAST has been analysed using a 0D global particle balance [71], estimating typical ion particle confinement 10 ms ≤ τ_p,I ≤ 100 ms [72] with a fuelling efficiency of 10%. A strong anticorrelation of τp,i∝nD2−0.7 with the neutral density in the tank is found. The wide range of variation suggests a genuine change in particle behaviour with increasing tank gas density. This is also observed in estimates of the energy confinement time at high density. Good particle control with a large tank volume can only be achieved with a baffled divertor chamber and fuelling from the divertor, the HFS or with pellets [71] as foreseen in MAST-U (see Section 12.3). The walls on MAST act as particle sinks throughout the discharge if conditioned by ⁴He glow discharge cleaning or boronisation, and in most discharges more than 80% of the inventory is retained in the walls [73].

Turbulent transport and plasma stability are strongly influenced by plasma rotation. Neutral beam driven future STs (eg, CTFs) are expected to have strong toroidal rotation due to the high torque from the beams, whilst ITER and DEMO rely on intrinsic rotation lacking momentum sources. It is important to understand the toroidal rotation profile in a tokamak. Momentum transport in L-mode on MAST is strongly linked to the ion energy transport and the data is generally in agreement with a Prandtl number P_ϕ = χ_ϕ/χ_i ≈ 1 in the core [58]. Only toward the edge a reduction of P_ϕ is seen supporting the presence of a momentum pinch by the so-called Coriolis drift [78]. The poloidal rotation in the core is found to be in agreement with neoclassical predictions [79].

The imaging of the filamentary structure of the ELM (see Fig. 12.9(b)) helped tremendously to understand the ELM crash [83], although the filamentary character of the ELM had been measured before [87,88]. Analysis of the fast videos allowed to determine the dynamics of the ELM crash. There are several types of bursty edge instabilities named ELMs that have been found in H-modes on tokamaks depending on the edge conditions [87]. At low input power above the H-mode threshold power type-III ELMs are found decreasing in frequency with increasing input power, associated with lower achievable p^ped. As the power is increased the confinement and the achievable p^ped increases, leading to type-I ELMs. The frequency of type-I ELMs increases with increasing power. Type-I ELMy H-mode is the standard operating regime foreseen for ITER. With higher shaping close to double null, at high collisionality very small ELMs occur. These type-II ELMs originate at the bottom of the pedestal [89] and on MAST co-exist with type-I ELMs at only slightly lower p^ped. Other types of ELMs have not been observed on MAST. Only for the type-I ELM the underlying instability has been identified [90]. The structures of type-I and type-III ELMs show no significant difference, whilst the structure of the type-II ELMs is fundamentally different [58,91]. The different ELM regimes on MAST can be distinguished in terms of pedestal parameters [58,92]. At high collisionality (Teped≤0.15keV,neped≥2.5·1019m−3) type-III ELMs exist. At lower collisionality (Teped≥0.15keV,neped≥2.5·1019m−3) these transit into type-I ELMs, whilst also a low-density branch of the type-III ELMs exists (Teped≥0.1keV,neped≤2·1019m−3), initially called type-IV ELMs sitting on the same isobar in Teped and neped as the high-density type-III ELMs. Due to the high-heat loads connected with type-I ELMs (Section 12.2.4) the smaller natural type-II ELMs with good confinement may be the more attractive regime for a fusion power plant, but this regime may not extrapolate to ITER, DEMO or future STs at lower collisionality. Type-III and type-IV ELMs also have acceptable heat load, but due to the lower pedestal pressure give lower confinement.

The prime candidate for the type-I ELM instability is the peeling-ballooning mode [90]. In general MAST pedestal profiles just before a type-I ELM are close to the stability boundary for peeling-ballooning modes [37,100,104,105], although some exceptions exist [36]. In contrast, the edge profiles for other ELM types are usually found well in the stable region. Stability analysis on its own is not enough to predict the pedestal performance in future devices. For this at least the width of the pedestal is needed. MAST data in general supports a scaling of the pedestal pressure width like Δp∝βpolped [58,92]. This is consistent with kinetic ballooning modes (KBM) clamping the pedestal gradient such that the width increases until the stability boundary is encountered [106]. Indeed the pedestal pressure and its gradient, eroded by the ELM crash, quickly recovers in a narrow part of the pedestal region and then the pedestal top pressure grows due to the widening of the transport barrier until the profiles become unstable (see Fig. 12.13). The pedestal width at low collisionality is much wider than at high collisionality. Distinguishing in the data set for the pedestal width between low and high collisionality shows that different effects could be important at the different collisionalities [94].

To maintain normal H-mode with type-I ELMs as an operating point for future devices the ELM instability has to be controlled in order to mitigate the transient heat loads to acceptable levels. Driven by the strong ITER needs and building on the strong research of the ELM and pedestal physics, a key area of research from 2008 onwards on MAST has been ELM mitigation by resonant magnetic perturbations. This has been facilitated by a versatile perturbation coil set consisting of 18 internal and 4 ex-vessel error field correction (EFC) coils. The in-vessel coil set has 12 coils below and 6 coils above the mid-plane. This allows application of n = 1–4 and 6 (n, toroidal mode number) perturbations with a flexible spectra of poloidal mode numbers m³ achieved by rotating the phase of the lower coil set with respect to the upper coil set.

Alfvén eigenmodes (AEs) are ubiquitous in plasma physics and are also encountered in astrophysical plasmas. In neutral beam heated plasmas on MAST a rich spectrum of AEs has been studied [26,117–121], improving the basic understanding of these instabilities. Both, ‘perturbative’ and ‘nonperturbative’ energetic particle modes are observed [26,118]. The ‘perturbative’ modes can be supported by the thermal plasma alone in the absence of fast particles and may be also driven unstable by an internal antenna (f_ant < 500 kHz,  I ≤ 10 A) [58,122] to measure their damping rates. For the ‘nonperturbative’ modes the dispersion properties are changed significantly by the presence of the fast-ion distribution itself leading to fast changes in the mode properties. At low β < 5% toroidal AEs (TAE) and elliptical AEs (EAE) are observed at f < 150 kHz. At medium 5% < β < 15% chirping and fishbone (n = 1, 2, 3) modes are encountered. For the TAE, EAE and chirping modes the amplitude and number of unstable modes decrease with increasing β [123] and at β > 15% the TAE and EAE activity vanish. At low density and high temperature, TAEs can have simultaneous up/down frequency sweeps δω/ω≤20% supporting a model for explosive TAEs [118,124] driven by a spontaneous hole-clump pair in the fast-ion distribution. Also chirping reversed shear Alfvén eigenmodes (RSAE) indicative of a nonmonotonic q-profile have been observed in discharges with a fast plasma current ramp-up (dI_p/dt ≈ 5 MA/s) or a double current ramp [26,119]. In addition to this rather low frequency f < 200 kHz activity, compressional AEs (CAE) have been observed in two bands between 0.4 MHz < f_CAE < 2.5 MHz [120]. In mixed H and D plasmas apart from CAEs also fast particle driven ion-ion hybrid waves with frequencies up to 2.25 times the central deuterium ion cyclotron frequency (f≤5MHz) have been observed [125]. The wave spectra of such plasmas in the ion cyclotron range should be similar to those of DT plasmas in tokamak reactors. With increasing H concentration, the waves in the D ion cyclotron range are suppressed. Analysis of the cut-offs and resonances for these waves using the cold plasma dispersion relation can explain the observed mode activity.

In order to understand the effect of AEs and other instabilities on the fast-ion distribution, direct measurements of the distribution function are needed. To characterise the confined six-dimensional fast-ion distribution f_FI(v,r), a suite of diagnostics (FID) has been built on MAST to measure different projections of f_FI in velocity and energy space as a function of radius [127]. Profiles of the fusion products, neutrons and 3 MeV protons, are measured with a scannable prototype collimated neutron camera (NC) [128] and a prototype charged fusion product detector [129] (CFPD), respectively. Profiles of trapped and passing fast ions are measured with a dual view fast-ion D_α diagnostic (FIDA) [130]. The interpretation of the measurements is aided by a suite of codes modelling the evolution of the fast-ion distribution combined with synthetic diagnostics [53,131,132], allowing a direct comparison with theory.

As will be discussed in Section 12.2.5, it is important in a spherical tokamak to avoid the q = 1 surface, since otherwise MHD is likely to reduce the performance. For this, current driven away from the magnetic axis is needed. The other advantage for driving current off axis is that a broader current profile leads to a lower internal inductance, l_i, allowing a higher elongation and therefore a higher normalised pressure, β_N, all of which will increase the performance of future STs and reduce in itself the need for driven current. The large vacuum vessel on MAST allows the comparison between off-axis and on-axis driven current even with neutral beams injecting at Z = 0 m. Since the flat top on MAST is not long compared to the current diffusion time, detailed transport analysis aided by good profile measurements including the current profile with MSE and the Z_eff profile are needed.

Turbulent transport is generally higher at the low-field side, due to destabilising curvature of the magnetic field. The high heat flux is transported in the SOL parallel to B to the high-field side (HFS). A direct evidence for this is the high ratio of power flowing to the outer and inner strike points of Γ_oi = 30–40 in DN L-mode [101] and Ohmic discharges [138]. In DN there is no parallel connection between inner and outer target. In H-mode, where turbulence is suppressed, the ratio reduces to about a factor of Γ_oi = 4–10 in quiescent inter-ELM periods, reflecting at the lowest ELM frequency the flux expansion between the inner and outer target [138–140]. This suggests a poloidally symmetric heat transport through the H-mode transport barrier. During the ELM Γ_oi rises again to the L-mode values [100,139]. In exact DN slightly more power flows in the ion ∇B-drift direction [101], though this balance changes during the ELM crash (see Section 12.2.2). During the crash also a spiral pattern is observed in the infrared (IR) images.

The ST geometry with its strongly varying field along the flux tube and the low field values at the LFS may influence the transport in the SOL. With respect to the parallel transport, indeed the mirror force is found to be important in MAST [150]. This force is proportional to ∇∥B/B, which is typically a factor of 10 larger in an ST compared to conventional tokamaks, due to the low aspect ratio. This term leads to changes in the charged particle velocity distributions near regions with large ∇∥B/B representing an effective, upstream particle and momentum source, which could drive strong upstream flows. Using coherence imaging, SOL flows have been visualised on MAST in the main chamber and the divertor using emission from C⁺, C²⁺ and He⁺ [149] (see Fig. 12.22).

As q_min approaches unity a saturated ideal (m,n) = (1,1) kink typically appears in MAST plasmas with early heating [133,162,163]. Equilibrium reconstructions constrained by field line angle measurements by MSE indicate that at the onset time of the mode the q-profile is slightly reversed in the core. Analytical theory for such q-profiles predicts that the ideal n = 1 kink should be unstable even at zero β below a critical Δq = q_min − 1. Indeed the equilibrium is found to be unstable to the n = 1 kink shortly before the observed onset and stable beforehand. The mode is preceded by fishbone instabilities that can exist without a q = 1 surface in a reversed q-profile. With Δq > 0.17 the fishbone drive dominates and below Δq < 0.17 the ideal kink mode dominates. The saturated kink degrades confinement, causes fast particle losses (see also Section 12.2.3) [127] and brakes the core plasma via neoclassical toroidal viscosity [162,163]. The mode saturates nonlinearly at a sufficiently large amplitude of the displacement where the stabilising field line bending term in the ideal MHD equations dominates and prevents further growth.

The sawtooth crash is a periodic relaxation of the core plasma pressure profile due to a partial or full reconnection of a core (m,n) = (1,1) kink. Frequent small crashes are advantageous, since impurities are expelled from the plasma core and may be needed to enhance He transport in a burning plasma. Infrequent, large crashes can be dangerous as they are likely to trigger unwanted instabilities (see Section 12.2.5.3), redistribute fast-ion loss (see Section 12.2.3) or even cause disruptions (see Section 12.2.5.4).

Tearing modes (TMs) are helical islands that form on rational q = m/n surfaces in the tokamak. These islands create a region of parallel connection between flux surfaces at different radii, reducing the achievable plasma pressure, which limits the performance of the fusion device. Large islands can also lead to disruptions (see Section 12.2.5.4), which need to be avoided. Typical q-profiles in a tokamak are stable against classical field line tearing, meaning that islands that are generated are healed. Even in such circumstances above a certain βpol=2μ0p/Bpol2 neoclassical effects connected to the field curvature can lead to a growth and saturation of the island – the NTM. For the NTM to form, a seed island is needed that can be generated by other MHD such as sawteeth, fishbone modes, large ELMs or error fields.

The study of disruptions is a key topic for future tokamaks. Firstly, the electromagnetic forces induced by the disruption can destroy the device. Secondly, the asymmetric heat loads on the plasma-facing components are likely to damage those surfaces. Thirdly, disruptions can create runaway electrons that can lead to extremely high, localised heat loads with the potential of destroying the vacuum vessel. Generally, disruptions on MAST are not a major threat to machine safety and the creation of a runaway beam has been rarely observed.

The routine plasma start-up on MAST is done using merging compression [31,179]. Here, the plasma breaks down around two internal P3 coils above and below the midplane (see Fig. 12.2) driven by capacitor banks. The plasma then expands toward the midplane attracted by the increasing current in each plasma ring. Closed flux surfaces are formed during a fast reconnection event, and finally the plasma is radially compressed toward the midplane, using the P4 and P5 coils, increasing the current further. The result of this plasma formation is a hot and dense plasma and currents of Ip≤0.5MA have been achieved routinely. Ion temperatures of T_i ≈ 1.2 keV in about 10 ms are achieved as the reconnection outflow during the merging phase thermalises through the fast shock and ion viscosity in the two downstream areas [180,181]. This temperature scales with the reconnecting poloidal field Bp∥2 and is about three times higher than measured for plasma formation using direct induction despite a factor of two smaller plasma current. Electron temperatures after the merging are of the order of T_e = 0.2 keV and centrally peaked. Clearly in-vessel coils are not feasible for future nuclear devices. Forming the plasma in two separate quadrupole nulls created above and below the midplane can achieve similar plasma currents. In this double-null merging schema the coils can be outside the vacuum vessel and experiments on MAST have achieved I_p = 370 kA with core temperatures and densities of T_e = 0.4–0.6 KeV and n_e(1.5–9) × 10¹⁹, respectively [61].

A very efficient method of noninductive start-up is provided using electromagnetic radio RF waves in the region of the electron cyclotron resonance (ECR). Unfortunately, ECR waves launched from the LFS with a frequency suitable for the low toroidal field on MAST will not propagate in the plasma at the densities of interest. Instead, these waves need to be converted into longitudinal electrostatic electron Bernstein waves. Such a conversion is possible close to the upper hybrid resonance if the waves are launched in a particular toroidal and poloidal angle and the density gradient is sufficiently high [15]. The conversion efficiency can be of the order of 100%. Current drive using EBW heating has been demonstrated on COMPASS-D [182] and using the same gyrotrons studied on MAST [15], but on MAST the conversion window at the available 60 GHz was too narrow to lead to definitive results. Indeed, frequencies of 19 GHz ≤ f_ECRH ≤ 30 GHz are better for the conditions encountered on MAST [183] (see also Section 12.3.3). The emission window for EBW heating has been studied using a novel synthetic aperture imaging diagnostic [184–186].