No abstract

While flowing liquid metal plasma-facing components (PFCs) represent a potentially transformative technology to enable long-pulse operation with high-power exhaust for fusion reactors, magnetohydrodynamic (MHD) drag in the conducting liquid metal will reduce the flow speed. Experiments have been completed in the linear open-channel LMX-U device [Hvasta et al., Nucl. Fusion. 58 (2018) 01602] for validation of MHD drag calculations with either insulating or conducting walls, with codes similar to those used to design flowing liquid metal PFCs for a Fusion Nuclear Science Facility [Kessel et al., Fusion Science Technol. 75 (2019) 886]. We observe that the average channel flow speed decreased with the use of conducting walls and the strength of the applied transverse magnetic field. The MHD drag from the retarding Lorentz force resulted in an increase of the liquid metal depth in the channel that ‘piled up’ near the inlet, but not the outlet. As reproduced by OpenFOAM and ANSYS CFX calculations, the magnitude and characteristics of the pileup in the flow direction increased with the applied traverse magnetic field by up to 120%, as compared to the case without an applied magnetic field, corresponding to an average velocity reduction of ~ 45%. Particle tracking measurements confirmed a predicted shear in the flow speed, with the surface velocity increasing by 300%, despite the 45% drop in the average bulk speed. The MHD effect makes the bulk flow laminarized but keeps surface waves aligned along the magnetic field lines due to the anisotropy of MHD drag. The 3D fringe field and high surface velocity generate ripples around the outlet region. It was also confirmed that the MHD drag strongly depends on the conductivity of the channel walls, magnetic field, and volumetric flow rate, in agreement with the simulations and a developed analytical model. These validated models are now available to begin to determine the conditions under which the ideal liquid metal channel design of a constant flow speed and fluid depth could be attained.

Summary In order to solve incompressible, nonideal magnetohydrodynamic (MHD) free‐surface flows, two weakly compressible smoothed particle hydrodynamics models, with and without the consideration of magnetic induction, are developed. The SPH formulation for magnetic induction magnetohydrodynamics (SPMHD), which is popular in astrophysical studies, is applied for the first time to incompressible free‐surface MHD flows, such as liquid metal flows, with the consideration of nonideal MHD effects and boundaries with arbitrary electric conductivity. An SPMHD implementation using the inductionless approximation is also proposed for both electrically conductive and insulating boundaries, in which a Poisson equation is solved to compute the Lorentz force instead of evolving the magnetic induction equation. Both proposed methods are validated against MHD benchmarks, including free‐surface MHD cases. The proposed inductionless SPMHD implementation has the advantages of stability and relaxed time‐step restrictions, but is only accurate at a low range of Hartmann numbers. For high Hartmann number problems, magnetic induction SPMHD model is more accurate. The computational efficiency and conservation error of the two models are compared and discussed.

This paper presents a simple and highly accurate method for capturing sharp interfaces moving in divergencefree velocity fields using the high-order Flux Reconstruction approach on unstructured grids. A well-known limitation of high-order methods is their susceptibility to the Gibbs phenomenon; the appearance of spurious oscillations in the vicinity of discontinuities and steep gradients makes it difficult to accurately resolve shocks or sharp interfaces. In order to address this issue in the context of sharp interface capturing, a novel, preconditioned and localized phase field method is developed in this work. The numerical accuracy of interface normal vectors is improved by utilizing a preconditioning procedure based on the level set method with localized artificial viscosity stabilization. The developed method was implemented in the framework of the multi-platform Flux Reconstruction open-source code PyFR [1]. Numerical tests in 2D and 3D conducted on different mesh types showed that the preconditioning procedure significantly improves accuracy. The results demonstrate the conservativeness of the proposed method and its ability to capture highly distorted interfaces with superior accuracy when compared to conventional and high-order VOF and level set methods. The high accuracy and locality of the proposed method offer a promising route to carrying out massively-parallel, high accuracy simulations of multi-phase, incompressible phenomena.

Divertor systems of fusion devices exhaust intense heat loads from the plasma, which degrades solid plasma-facing components (PFCs). Fast liquid metal (LM) flow divertors may be more advantageous for this purpose. However, LMs have risk of piling due to intense magnetohydrodynamic (MHD) drag. Despite this, severe deceleration of the flow could be countered with the injection of currents that are transverse to external magnetic fields, allowing to thrust the flow with jxB forces. The injection of currents as an approach to propel LM-divertor flows has remained experimentally understudied. This article focuses on the evaluation of jxB-thrust and finding its drawbacks. This paper evidences that the simple operation of a LM-flow divertor with jxB-thrust, without any of the instabilities caused from reactor plasmas or parasitic currents, already presents intrinsic challenges.

jxB-thrust was experimentally tested with free-surface-LM flows, a vertical magnetic field and and externally applied current. Experiments were reviewed with a theoretical model, showing agreement in the trends of theory and experiments. Full 3D-MHD-free-surface flow simulations were also performed with FreeMHD and confirmed the sensitivity to unstable flow behavior in LM systems when applying external currents. Furthermore, excessive power requirements are expected for the implementation of jxB-thrust at the reactor scale, making these systems inefficient for commercial devices.