SPECTRE: A robust solver for 3D equilibria with arbitrary topology

summary

The gist

SPECTRE presents a new equilibrium code based on the Multi-Region relaxed MHD model for robustly calculating 3D equilibria with general magnetic topology allowing for flux surfaces…

In short

SPECTRE is a new solver for 3D plasma equilibria using a Multi-Region relaxed MHD model. It improves upon previous solvers by using trust-region based minimization and real-space force representation on an interface grid. This allows SPECTRE to robustly calculate equilibria with complex magnetic topologies, such as flux surfaces, islands, and chaos.

Key concepts

Multi-Region relaxed MHD (MRxMHD)
This is a variational model used to describe general plasma equilibria. It is based on the relaxation principle where plasma evolves to minimize energy while conserving global helicity. The model uses an energy functional composed of magnetic energy and helicity for each volume.
Trust-region based minimization
SPECTRE uses this method to find equilibrium states by minimizing the force in real space. It works by locally approximating the objective function with a quadratic model, ensuring that optimization steps remain bounded within a defined domain. This makes the solver more robust and reliable than root-finding methods.
Real space force representation
Instead of using Fourier descriptions, SPECTRE represents the magnetic force in real space on a chosen grid that samples an interface. This allows for advanced sampling schemes, such as denser sampling where curvature is high, providing a more accurate and scaled force representation for the solver.

Terminology used across episodes

This episode discusses

The paper

SPECTRE: A robust solver for 3D equilibria with arbitrary topology · Read on arXiv

E. Balkovic, J. Loizu, E. Lanti, C. Smiet, C. Lazzati, Ramasamy, A. Goodman, J. Geiger

École Polytechnique Fédérale de Lausanne Institute for Plasma Physics, Swiss Plasma Center, Switzerland Institute of SCITAS, École Polytechnique Fédérale de Lausanne, Switzerland Proxima Fusion GmbH Munich Max-Planck-Institut für Plasmaphysik Greifswald York Plasma Institute University of York

We present SPECTRE, a new equilibrium code based on the Multi-Region relaxed MHD model for robustly calculating 3D equilibria with general magnetic topology, allowing for flux surfaces, magnetic islands, and chaos. The code builds on a previous MRxMHD solver, SPEC, but performs significantly better thanks to a new formulation of force, the use of a stable trust-region-based least squares minimization scheme, as well as several additional features. SPECTRE is verified through application to configurations with known equilibrium solutions, in vacuum and with finite beta, both in the fixed boundary and the free boundary mode. Notably, these include vacuum equilibria of a quasi-axisymmetric (QA) device in fixed-boundary mode, and of W7-X in the free-boundary mode, along with a classical stellarator finite-beta free-boundary case. Finally, the solver is applied to a modern optimized finite-beta quasi-isodynamic (QI) configuration, where we demonstrate calculation of a strongly-shaped equilibrium with a core island which is in agreement with a HINT calculation.

DOI: 10.1088/1361-6587/aebef8

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "SPECTRE: A robust solver for 3D equilibria with arbitrary topology".

Mira: Here's the gist of the paper. SPECTRE presents a new equilibrium code based on the Multi-Region relaxed MHD model for robustly calculating 3D equilibria with general magnetic topology allowing for flux surfaces, magnetic islands, and chaos.

Kai: General 3D Equilibria Background Time-independent solutions to the Magnetohydrodynamic (MHD) equations satisfying J × B = ∇p are essential for the description of magnetically confined fusion plasmas.

Mira: In tokamaks, where the plasma possesses axisymmetry, the magnetic field foliates a continuous set of nested flux surfaces.

Kai: However, axisymmetry does not hold in general, e.g., in stellarators or tokamaks with resonant magnetic perturbations.

Title and authors: Kai: So we’re talking about this paper, SPECTRE: A robust solver for three dee equilibria with arbitrary topology, and the authors are Balkovic, Loizu, Lanti, Smiet, Lazzati, Ramasamy, Goodman and Geiger <ref:2607.27135#pg1,SPECTRE: A robust solver for 3D equilibria with arbitrary topology>.

Mira: The title itself tells you the goal is to create a new equilibrium code that can handle three dee magnetic fields in a very general way <ref:2607.27135#pg1>. It's not just for simple shapes anymore; it includes flux surfaces but also magnetic islands and even chaos.

Lev: So what this means practically is that we are moving away from solving problems where we assume the field is perfectly symmetric or has simple nested surfaces, which was the norm for a long time in tokamak simulations.

Kai: Exactly, and they state that general three dee equilibria are really important because they leave out potential equilibrium states with more general topology than what simpler solvers can find <ref:2607.27135#pg1>.

Mira: They separate these into two categories: relaxation codes that evolve the plasma like a fluid, and the Multi-Region relaxed MHD or MRxMHD approach which directly finds an equilibrium state as a set of Taylor-relaxed plasma subvolumes.

Lev: The MRxMHD model itself is based on the relaxation principle devised by Taylor in one thousand nine hundred seventy-four where you minimize the energy while keeping global helicity conserved <ref:2607.27135#pg2>.

Kai: And SPECTRE takes that variational model, SPEC, and makes it more robust by using a new force formulation and better minimization techniques.

Mira: The core idea is that they are using a global polynomial-Fourier basis to express the interface geometry in cylindrical coordinates, which lets them represent the interface shape in terms of poloidal and toroidal angles theta and zeta.

Lev: That representation is not unique unless you impose extra angle constraints, which means they have to be careful about what kind of boundary shapes they are testing.

The paper's summary: Kai: So the paper summarizes how SPECTRE solves these general three dee equilibria by using that MRxMHD model, expressing the geometry via a double Fourier series for interfaces <ref:2607.27135#pg1>.

Mira: They explain that to solve this, SPEC requires knowing the total toroidal and poloidal fluxes enclosed by each subvolume along with the total subvolume helicity.

Lev: And they show how you can iterate on those helicity and flux values using various combinations of constraints like toroidal current or integrated current in each volume.

Kai: They also mention that SPEC previously struggled with things like δ-function current sheets on resonant rational surfaces and singular Pfirsch-Schlüter current densities when pressure gradients are present, which can cause numerical issues.

Mira: SPECTRE is designed to handle those issues by using the trust-region minimization and the real space force representation to avoid those specific numerical pitfalls.

Lev: They claim that SPECTRE can recover correct fixed boundary equilibria with fully nested flux surfaces and without islands when it's in that simple case, agreeing with results from VMEC.

Kai: So, in summary, the paper shows they took an existing variational model and significantly improved the solving process to handle a much wider range of magnetic field topologies than before.

The paper's improvements: Mira: The key improvements they highlight are three things: first, the trust-region based minimization for force calculation which guarantees bounded steps.

Lev: So it’s not just a speedup; it’s about robustness. It avoids the issues you see with methods like the Powell method because those steps stay within a predictable domain.

Kai: Second, they switched to a real space representation of the force on a chosen grid, which lets them sample interfaces much better in high-curvature regions.

Mira: That means they get a better scaled force representation compared to just using truncated Fourier descriptions for that geometry.

Lev: And third, coupling it with map2disc so it can use isosurfaces of the Laplace equation as an initial guess for the MRxMHD interfaces.

Kai: So, that combination—better force minimization, better interface sampling, and a smarter initialization—is what really allows SPECTRE to be more robust in finding those complex states.

Conclusion: Mira: To wrap up the paper on SPECTRE: A robust solver for three dee equilibria with arbitrary topology, they show that this new code can reliably calculate equilibria across a much broader range of magnetic field topologies <ref:2607.27135#pg1,SPECTRE: A robust solver for 3D equilibria with arbitrary topology>.

Kai: They verified it on configurations with known solutions in both vacuum and finite beta, showing agreement with HINT calculations for strongly-shaped equilibria.

Lev: And they confirmed that when things are simple—like fully nested flux surfaces without islands—SPECTRE agrees with VMEC results.

Kai: The implication is that we have a more reliable tool now to explore the space of possible magnetic field configurations, which is really important for designing fusion devices.

Mira: It opens the door to studying those complex, non-axisymmetric scenarios where magnetic islands and chaos are present, which were previously too difficult to tackle reliably.

Lev: For me, I'm just glad they showed how the overhead of these new solvers lands; it seems like a worthwhile investment for getting access to that more general solution space.

Kai: Yeah, SPECTRE is now out there publicly available under the MIT license at gitlab.com/spectre-eq/spectre, so other researchers can try it out with their own setups.

Mira: It's a solid step forward in computational physics, moving beyond just solving for simple tokamak geometries to tackling the more general three dee problem of magnetic confinement <ref:2607.27135#pg1>.

More episodes

← Home