Machine learning (ML)-driven reduced-order modelling is applied to accelerate steady-state convergence in three-dimensional, nonlinear, flux-driven two-fluid simulations of boundary plasma turbulence. A parametric scan of plasma resistivity, heating and density sources is performed to generate comprehensive datasets across various turbulent regimes for model training and validation. To efficiently manage and interpret these datasets, we apply the proper orthogonal decomposition technique to reduce the dimensionality of key plasma quantities such as plasma density, temperature, electric potential and vorticity. Data-driven models are trained to map physical parameters to low-dimensional representation, enabling the rapid generation of quasi-steady-state plasma profiles. The results demonstrate that density, temperature and electric potential are qualitatively well captured with a relatively low number of bases, whereas vorticity requires a larger number of bases due to its fine spatial structures. A comparison between ML-generated restarts and simulations from scratch demonstrates a significant computational advantage of the ML approach, reducing simulation time by up to a factor of three. This hybrid framework, combining data-driven reduced-order modelling with first-principles simulations, highlights the potential of ML to accelerate plasma turbulence modelling, making high-fidelity simulations more computationally feasible for large-scale fusion devices, such as ITER and DEMO.

The effects of negative triangularity (NT) on boundary plasma turbulence in double-null (DN) configurations are investigated using global, nonlinear, three-dimensional, flux-driven two-fluid simulations. Negative triangularity plasmas exhibit suppressed interchange-driven instabilities, resulting in enhanced confinement and lower fluctuation levels compared with positive triangularity (PT) plasmas. This reduction in interchange instability is associated with the weakening of curvature effects in the unfavourable region, caused by the stretching of magnetic field lines at the outer midplane. The magnetic disconnection between the turbulent low-field side and the quiescent high-field side results in most of the heat flux reaching the DN outer targets. In NT plasmas, the power load on the outer target is reduced, while it increases on the inner target, indicating a reduced in–out power asymmetry compared with PT plasmas. Furthermore, the analysis of power load asymmetry between the upper and lower targets shows that the absolute magnitude of up–down power asymmetry is mitigated in NT plasmas, mainly due to the reduced total power crossing the separatrix. The reduction of interchange instabilities in NT plasmas also affects the blob dynamics. A three-dimensional blob analysis reveals that NT plasmas feature smaller blob sizes and slower propagation velocities. Finally, an analytical scaling law for blob size and velocity that includes plasma shaping effects is derived based on the two-region model and is found to qualitatively capture the trends observed in nonlinear simulations.

Runaway electrons (REs), generated during plasma disruptions in tokamaks, pose significant challenges due to the risk of causing damage to the first wall of a device. Understanding the interaction between REs and magnetohydrodynamic (MHD) instabilities is crucial for predicting a safe operation of large future tokamak devices in which RE generation will be drastically enhanced due to the high plasma current. In this work, we introduce a hybrid fluid–kinetic model within the three-dimensional nonlinear MHD code JOREK (Hoelzl et al. 2021 Nucl. Fusion, vol. 61, 065001; 2024 Nucl. Fusion, vol. 64, 112016), treating REs kinetically using a relativistic guiding-centre approach, while describing the background plasma by ansatz-based reduced MHD equations. At first, comprehensive benchmark studies are conducted regarding the two-dimensional equilibrium force balance with $J_{total}= J_{RE}$, and the linear stability of three-dimensional tearing modes (TMs), verifying the accuracy of the model against analytical predictions and other numerical methods, e.g. the full-orbit approach in JOREK and the fluid model in M3D-C1. These benchmark studies build a solid foundation for applying our model to more complex nonlinear scenarios. In this respect, we confirm that the nonlinear saturation of TMs is significantly influenced by the presence of REs. Previous analytical studies (Helander et al. 2007 Phys. Plasmas vol. 14, 122102) suggest that in the case of small $\varDelta ^\prime$, the saturation width of the magnetic island driven by REs is roughly 1.5 times larger than in the otherwise identical Ohmic current scenario. Our simulations are quantitatively in line with this prediction. Moreover, REs alter the energy evolution within the magnetic reconnection process and decouple the bulk plasma and magnetic fields. In summary, RE-driven magnetic reconnection leads to larger magnetic islands but weaker reconnection flows.

Neutral atoms recycled from wall interaction interact with confined plasma, thereby refuelling it, most strongly in the region closest to the wall. This occurs near the X-point in diverted configurations, or else near the wall itself in limited configurations. A progression of analytic models is developed for neutral density in the vicinity of a planar or linear source in an ionising domain. First-principles neutral transport simulations with DEGAS2 are used throughout to test the validity and limits of the model when using equivalent sources. The model is further generalised for strong plasma gradients or the inclusion of charge exchange. An important part of the problem of neutral fuelling from recycling is thereby isolated and solved with a closed-form analytic model. A key finding is that charge exchange with the confined plasma can be significantly simplified with a reasonable sacrifice of accuracy by treating it as a loss. The several assumptions inherent to the model (and the simulations with which it is compared) can be adapted according to the particular behaviour of neutrals in the divertor and the manner in which they cross the separatrix.

Recently, Alfvénic ion temperature gradient (AITG) modes have been observed in the core plasma on the HL-2A tokamak. Only when electron cyclotron resonance heating (ECRH) and neutral beam injection are simultaneously injected into the deuterium plasma do the AITG modes become unstable. The instability is electromagnetic and localised in the core plasma with an internal transport barrier. Dynamic evolution of AITG modes is greatly affected by the off-axis ECRH. Theoretical analysis suggests that there is a strong dependence of the AITG modes on $\eta _i\simeq \boldsymbol{\nabla }\ln T_i/\boldsymbol{\nabla }\ln n_i$, where $n_i$ is the ion density. It is also found that ECRH can enhances AITG modes by causing a drop of electron density and an increase of $\tau =T_e/T_i$; here $T_e$ and $T_i$ are the electron and ion temperatures, respectively. Besides, high-power ECRH may also change the safety factor or magnetic shear and then contribute to the mitigation of AITG modes. The new findings can not only enrich scientific knowledge for pressure gradient-driven instability, but also be beneficial to active control of core-localised electromagnetic modes in future fusion devices.