Computation of generalised magnetic coordinates asymptotically close to the separatrix

S. Benjamin, N.C. Logan, C. Hansen · 2025

Read Computation of generalised magnetic coordinates asymptotically close to the separatrix by S. Benjamin, N.C. Logan, C. Hansen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-04934.txt.

Introduction

Stuart Benjamin, Nikolas Logan, and Christopher Hansen compute magnetic coordinates right up to the edge of a tokamak plasma. Their twenty twenty-five paper serves diverted tokamaks, where the last closed flux surface ends at an X-point. Field line sums blow up near that point, breaking standard tools.

New integrals give generalized coordinates from a given flux map, valid ever closer to the separatrix. The method is built into the GPEC and DCON code suite used for plasma stability work. It lets spectral 3D codes model edge behavior with care. A robust edge map is a first step toward predicting how the plasma rim will act in real machines.

Catalog record

CREATORS. Benjamin, N.C. Logan, C. Hansen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-04934

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Computation of generalised magnetic coordinates asymptotically close to the separatrix S. Benjamin, N.C. Logan, C. Hansen Abstract Integrals to calculate generalised magnetic coordinates from an input magnetic flux function asymptotically close to the separatrix are presented, and implemented in the GPEC/DCON code suite. These integrals allow characterisation of the magnetic equilibrium of a diverted tokamak, in magnetic coordinates, arbitrarily close to the last closed flux surface, avoiding the numerical issues associated with calculating diverging field line integrals near a magnetic x-point. These methods provide an important first step in the development of robust asymptotic equilibrium behaviour for spectral 3D MHD codes at the separatrix. Preprint: AIP/123-QED I Introduction Magnetic coordinates have provided myriad advantages in magnetic confinement fusion plasma research. These coordinates, which adhere to the magnetic geometry of the device, can vastly simplify both the analytical expression of stability problems and their computation. Their application in diverted tokamaks, however, is hindered by the fact that the poloidal magnetic field approaches zero at the magnetic x-point on the last closed flux surface (LCFS). This null point in the poloidal field produces an unavoidable divergence in one of two periodic magnetic angles in tokamak magnetic geometry. The result is that the computational domain of spectral MHD codes must be truncated somewhere inside of the LCFS, and unfortunately global stability calculations retain some degree of sensitivity to the exact point of truncation. In this publication we will discuss this problem in some detail, before presenting a first step to its final solution: reliable and simple analytic integrals that extend the computational domain arbitrarily close to the magnetic separatrix. These integrals have been implemented in the GPEC/DCON code suite [ 4 , 21 , 23 ] . Note the complete resolution of the spectral edge truncation problem requires the development of appropriate asymptotic numerical methods at the separatrix, that will depend on the particular stability problem being solved. II Background The magnetic coordinate representation we will apply is the same as that used in DCON [ 4 ] . We repeat its formulation here for convenience. Let us assume the existence of a magnetic flux coordinate \psi (as is appropriate in axisymmetry [ 3 ] ) such that \mathbf{B}\cdot\nabla\psi=0 . Due to the divergence-free property of the magnetic field, we can define two magnetic angle coordinates, \zeta and \theta that are periodic around the toroidal and poloidal directions of the tokamak respectively, such that field lines are straight lines in the \zeta,\theta plane, \mathbf{B}=(\nabla\zeta-q\nabla\theta)\times\nabla\psi, (1) with the gradient of the field lines given by the q-profile, q(\psi)\equiv\frac{\mathbf{B}\cdot\nabla\zeta}{\mathbf{B}\cdot\nabla\theta}. The standard stability problem through which we will discuss edge truncation is that of finding the solution for minimum-energy ideal linear perturbations in axisymmetric geometry. To refresh the reader’s memory on this problem, consider the energy of a general ideal internal plasma perturbation about an equilibrium state [ 3 ] \mathbf{J}\times\mathbf{B}=\nabla p : Figure 1: DCON calculation of the n=1 , minimum- \delta W plasma displacement in a single-null DIII-D equilibrium for two different edge truncations. Here \psi_{edge}=0.987 and 0.992 for panels a and b respectively (3 s.f.). The q_{edge}=4.93 displacement is highly localised at the plasma edge, forming an external kink, while the q_{edge}=5.21 displacement has greater internal structure. \displaystyle\delta W=\frac{1}{2\mu_{0}}\int_{V}d\mathbf{R}[ \displaystyle\mathbf{B_{1}}^{2}+\mathbf{J}\cdot\bm{\xi}\times\mathbf{B_{1}}+ \displaystyle\mu_{0}(\bm{\xi}\cdot\nabla p)(\nabla\cdot\bm{\xi})+\mu_{0}\gamma p(\nabla\cdot\bm{\xi})^{2}] (2) where \mathbf{B_{1}}=\nabla\times(\bm{\xi}\times\mathbf{B}) is the perturbed magnetic field, \bm{\xi}=\frac{\partial\mathbf{R}}{\partial\psi}\xi_{\psi}+\frac{\partial\mathbf{R}}{\partial\zeta}\xi_{\zeta}+\frac{\partial\mathbf{R}}{\partial\theta}\xi_{\theta} is a small, general plasma displacement in contravariant form, \mathbf{J},\mathbf{B} and p are the background current, magnetic field, and pressure respectively, and \gamma is thermodynamic ratio of specific heats. Following [ 4 ] , we can write the contravariant coefficients of \bm{\xi} in terms of a Fourier basis in magnetic angles: \xi_{j}(\psi,\zeta,\theta)=\sum_{m=-\infty}^{\infty}\sum_{n=-\infty}^{\infty}\bar{\xi}_{j}|_{m,n}(\psi)e^{2\pi i(m\theta-n\zeta)},\hskip 5.69054ptj=(\psi,\zeta,\theta) where m and n are poloidal and toroidal mode numbers respectively. After substituting the Fourier representation of \bm{\xi} into Eq. 2 , and minimising \delta W by dropping strictly positive terms (eliminating \xi_{\theta} from the integral & rendering the perturbations incompressible [ 3 ] ), one can rewrite \xi_{\zeta} in terms of \xi_{\psi} to arrive at, \displaystyle\delta W \displaystyle=\frac{1}{2\mu_{0}}\int d\psi[\bm{\Xi}^{\prime{\dagger}}\mathbf{F}\bm{\Xi}+\bm{\Xi}^{\prime{\dagger}}\mathbf{K}\bm{\Xi}+\bm{\Xi}^{\dagger}\mathbf{K}^{\dagger}\bm{\Xi}^{\prime}+\bm{\Xi}^{\dagger}\mathbf{G}\bm{\Xi}], where \bm{\Xi} is a vector of Fourier amplitudes of \xi_{\psi} only: \displaystyle\bm{\Xi} \displaystyle=\{\bar{\xi}_{\psi}|_{m,n}(\psi)|m_{\rm low}\leq m\leq m_{\rm high}\}, (3) and \mathbf{F},\mathbf{G} and \mathbf{K} are matrices of Fourier components of equilibrium quantities [ 4 ] . We apply the Euler Lagrange equations to the above integral to get a second order matrix ODE: \displaystyle\mathbf{L}\bm{\Xi}(\psi)\equiv-(\mathbf{F}\bm{\Xi}^{\prime}+\mathbf{K}\bm{\Xi})^{\prime}+(\mathbf{K}^{\dagger}\bm{\Xi}^{\prime}+\mathbf{G}\bm{\Xi})=0. (4) Eq. 4 describes the minimum-energy, incompressible perturbations in an axisymmetric magnetic field. It demonstrates how magnetic coordinates have reduced what could have been a very complex 2 or 3D problem (in cylindrical or Cartesian coordinates, for example) to a 1D integration that can be solved using pre-existing adaptive integrators such as LSODE [ 10 ] . Applications of Eq. 4 include efficiently testing internal mode ideal stability via the sign of a 1D scalar [ 4 ] , calculating the total plasma + vacuum \delta W [ 4 , 7 ] , calculating the least stable mode structure for resistive wall mode control [ 2 ] , calculating plasma response to resonant magnetic perturbation and/or error field correction coils [ 21 , 23 ] , calculating the corresponding neoclassical transport and/or kinetic-MHD stability [ 18 , 24 ] , and calculating the ideal-region drive to tearing by examining ratios of divergent & non-divergent components of ideal solutions at rational q surfaces [ 26 , 5 , 6 ] where q(\psi)=m/n . To fully understand the difficulty of edge truncation, the reader must understand that Eq. 4 becomes singular at rational surfaces, due to the incompressibility condition [ 3 ] . To illustrate this, \mathbf{F},\mathbf{G} and \mathbf{K} can be written in terms of a diagonal matrix \mathbf{Q} : \mathbf{F}=\mathbf{Q\bar{F}Q}\text{, }\text{ }\mathbf{K}=\mathbf{Q\bar{K}}\text{, }\text{ }\mathbf{G}=\mathbf{\bar{G}}\text{, }\text{ }\text{ }\text{ }\text{ }\mathbf{Q}_{m,m^{\prime}}\equiv(m-nq(\psi))\hskip 0.85358pt\delta_{m,m^{\prime}}, such that when Eq. 4 is reformulated as two coupled first order ODEs [ 4 ] : \displaystyle\mathbf{u}^{\prime} \displaystyle=\mathbf{H}\mathbf{u}\text{, where }\mathbf{u}=\begin{pmatrix}\bm{\Xi}\\ \mathbf{F}\bm{\Xi}^{\prime}+\mathbf{K}\bm{\Xi}\end{pmatrix}, (5) \displaystyle\mathbf{H} \displaystyle=\begin{pmatrix}-\mathbf{F^{-1}}\mathbf{K}&\mathbf{F^{-1}}\\ \mathbf{G}-\mathbf{K}^{\dagger}\mathbf{F}^{-1}\mathbf{K}&\mathbf{K}^{\dagger}\mathbf{F}^{-1}\end{pmatrix} \displaystyle=\begin{pmatrix}-\mathbf{Q^{-1}}\mathbf{\bar{F}^{-1}}\mathbf{\bar{K}}&\mathbf{Q^{-1}}\mathbf{\bar{F}^{-1}}\mathbf{Q^{-1}}\\ \mathbf{\bar{G}}-\mathbf{\bar{K}}^{\dagger}\mathbf{\bar{F}}^{-1}\mathbf{\bar{K}}&\mathbf{\bar{K}}^{\dagger}\mathbf{\bar{F}}^{-1}\mathbf{Q^{-1}}\end{pmatrix}, where Figure 2: DCON calculation of \delta W_{n=1} for a low q95 H-mode pilot-plant scenario generated in TokaMaker [ 9 ] . \psi is normalised poloidal flux. \mathbf{Q}^{-1}=\begin{pmatrix}\frac{1}{m_{\rm low}-nq(\psi)}&0&\ldots&0\\ 0&\frac{1}{(m_{\rm low}+1)-nq(\psi)}&&0\\ \vdots&&\ddots&\vdots\\ 0&0&\ldots&\frac{1}{m_{\rm high}-nq(\psi)}\end{pmatrix}. The rational surface singularity becomes clear in that \mathbf{Q^{-1}} will vanish when q(\psi)=m/n . Furthermore, near some singular surface \psi_{r} where m_{r}-n_{r}q(\psi_{r})=0 , the resonant Fourier perturbation amplitude \bar{\xi}_{\psi}|_{m_{r},n_{r}} takes on two asymptotic forms: a diverging energy solution and a finite energy solution. In the ideal stability problem, the diverging energy solution is set to zero [ 4 ] . Figure 3: Figure reproduced with permission from [ 25 ] , showing the n=2 edge plasma response of \delta W and |B_{p1}| as equilibrium truncation q-value ( q_{a} ) is varied for two DIII-D equilibria. ‘Ref 36 Eq’ refers to the second equilibrium, initially analysed in [ 17 ] . \psi_{N} refers to normalised poloidal flux. HFS and LFS refer to the high-field and low-field sides of the plasma respectively. Despite the usefulness of magnetic coordinate formulations, they suffer a key setback in that the angle coordinates are undefined at the separatrix. This is because in a diverted tokamak plasma with a separatrix, the poloidal component of the magnetic field is zero at the x-point. The magnetic field lines do not close in the poloidal plane, and the periodic magnetic angle cannot be defined.

Read the complete text