Stochastic single-stage stellarator optimization using fixed-boundary equilibria Publication type: Paper
Pedro F. Gil 1,∗, Jason Smoniewski 2,3, Rogerio Jorge 4,5, Paul Huslage 1 and Eve V. Stenson 1 · 2026
Read Stochastic single-stage stellarator optimization using fixed-boundary equilibria Publication type: Paper by Pedro F. Gil 1,∗, Jason Smoniewski 2,3, Rogerio Jorge 4,5, Paul Huslage 1 and Eve V. Stenson 1 (2026) 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-2603-11699.txt.
Introduction
Gil and co-authors merge single-stage stellarator design with stochastic coil optimization to harden the device against coil errors. The plasma boundary, from an MHD solver in fixed-boundary mode, links to randomly perturbed coils through the squared flux. That setup steers the search away from sharp local minima toward sturdier shapes. Two new designs come out: one quasi-axisymmetric and one quasi-helically symmetric. Under later coil perturbations, they beat standard methods on squared flux, quasisymmetry, and particle loss.
Catalog record
| CREATOR | Pedro F. Gil 1,∗, Jason Smoniewski 2,3, Rogerio Jorge 4,5, Paul Huslage 1 and Eve V. Stenson 1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52240 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.11699
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stochastic single-stage stellarator optimization using fixed-boundary equilibria Publication type: Paper Pedro F. Gil 1,∗, Jason Smoniewski 2,3, Rogerio Jorge 4,5, Paul Huslage 1 and Eve V. Stenson 1 Abstract In this paper, single-stage stellarator optimization is combined with stochastic coil optimization to improve the robustness of the stellarator as compared to deterministic methods. The plasma boundary, solved with an MHD solver in fixed-boundary mode, is linked to a set of randomly perturbed coils via the squared flux. The optimizer avoids sharp local minima and can reach improved configurations. Two different configurations obtained with our method, one quasi-axisymmetric and one quasi-helically symmetric, are compared against both the standard stochastic stage II method and the single-stage method. The new configurations shown here yield improved squared flux, quasisymmetry, and particle loss following a posteriori perturbation of the coils. 1 Introduction In recent years, interest in stellarators as a viable option for achieving commercial fusion has increased. Stellarators are toroidal devices that confine a plasma using 3D-shaped coils. These coils generate a magnetic field with a rotational transform without significant current in the plasma and allow for steady-state operation. Due to relaxing the requirement for axisymmetry, stellarator optimization has broad potential to meet multiple physics requirements, such as minimizing neoclassical transport, mitigating MHD instabilities, and improving the confinement of fast particles [ 4 , 9 , 18 ] . One main disadvantage of stellarators is their complexity to build. This also means that it is difficult to meet building tolerances. A notable example is the National Compact Stellarator eXperiment (NCSX) in the USA, which was canceled primarily because of difficulty meeting the tolerances [ 24 ] . Tolerances here refer to the allowed manufacturing and assembly geometrical deviations of the coils. Moreover, although the construction of Wendelstein 7-X has proven to be a success, tolerances were relatively small compared to the size of the coils: 2 mm allowed deviations for coils with 3.5 m diameter. These were determined to keep the relative magnetic field deviations within 10^{-4} [ 22 , 1 ] . These results have sparked extensive work on ways to improve the shape of the non-planar coils and design plasma shapes that simultaneously meet physics requirements and relax engineering constraints [ 13 , 14 , 6 , 8 , 15 , 18 , 17 , 21 ] . Recent work has shown that it is possible to reduce neoclassical transport in stellarators across the volume and achieve values similar to those of axisymmetric devices by targeting quasisymmetry, which acts as a proxy for neoclassical transport [ 18 ] . Quasisymmetry is a property of the magnetic field amplitude that, if invariant along a given Boozer coordinate, allows for the confinement of collisionless trapped particles. This method was first used in a stage I approach, where solely the plasma equilibrium is optimized. Typically, this is followed by a stage II optimization, where a magnetostatic inverse problem is solved to determine coils that produce a field matching the previously-found equilibrium surface. The main disadvantage of this sequential two-stage approach is that the equilibrium optimization is done ’blindly’. It often results in plasmas that are far too complex to find buildable coils that accurately reproduce the plasma shape. Moreover, the inverse problem of finding coils is ill-posed, as the number of coil degrees of freedom is far larger than the number of equations that are available to regularize the coils. The consequence is that for one equilibrium, countless different coil sets can be found. This makes the entire optimization procedure complex. Two notable developments in this domain are single-stage methods [ 10 , 7 , 13 ] and stochastic coil optimization [ 26 , 25 , 19 ] . The former combines the shaping of the plasma while simultaneously modifying the coils. This is needed since stage II, although effective at meeting the coil engineering constraints, such as maximum curvature, length, or torsion, it does so in detriment of field accuracy, due to poor compatibility between plasma and coils. Additionally, stochastic optimization has been developed to address the challenges associated with engineering tolerances. The optimization targets a cloud of perturbed coils surrounding the central coil set, rather than a single coil. For a “narrow” minimum, the cloud of perturbed coils will increase the value of the optimization cost function, as compared to a “broader” minimum. This means that, under coil perturbations, the value of the cost function should not increase as much as for configurations optimized with traditional methods, resulting in a more robust stellarator. The current paper combines the advantages of both these methods. This article is organized as follows: In Section 2 , the individual methods of single-stage and stochastic coil optimization are briefly explained, and then the merging of the two is presented. In Section 3.2 , the stochastic single-stage optimized configurations are presented and compared to the deterministic version. Moreover, two stellarator configurations, one quasi-axisymmetric (QA) and one quasi-helically symmetric (QH), were optimized and checked against quasisymmetry, coil accuracy, robustness, and alpha particle confinement metrics. We conclude with Section 4 . 2 Optimization Methods 2.1 Single Stage with Fixed Boundary Equilibrium In the single-stage approach to stellarator optimization, the shape of both plasma and coils are optimized simultaneously. The method adopted in this paper was developed by R. Jorge et al. [ 13 ] with the optimization software SIMSOPT [ 16 ] , where the plasma is represented via a magnetohydrodynamic (MHD) equilibrium from the Variational Moments Equilibrium Code (VMEC) [ 11 ] . VMEC solves for ideal MHD equilibria by minimizing the MHD potential energy W on a series of toroidally nested flux surfaces via a steepest descent method, where W=\int\left(\frac{p}{\gamma-1}+\frac{|\vec{B}|^{2}}{2\mu_{0}}\right)dV. (1) Here p is the plasma pressure, B the magnetic field, \gamma the adiabatic index, and V the volume considered. For a free-boundary problem there is no prescribed boundary, instead the coils and their currents are given, then an iterative process is used to find the correspondent plasma boundary. While it is more physically accurate, a free-boundary problem is more computationally expensive and more difficult to converge. For a fixed-boundary problem, the computational domain is given by the LCFS, provided as an input. This work uses VMEC to solve fixed-boundary problems, where the surface is parameterized in cylindrical coordinates via: \displaystyle R(\theta,\phi)=\sum_{m=0}^{M_{pol}}\sum_{n=-N_{tor}}^{N_{tor}}\left[a_{m,n}\cos(m\theta-n_{fp}n\phi)+b_{m,n}\sin(m\theta-n_{fp}n\phi)\right], (2a) \displaystyle Z(\theta,\phi)=\sum_{m=0}^{M_{pol}}\sum_{n=-N_{tor}}^{N_{tor}}\left[c_{m,n}\cos(m\theta-n_{fp}n\phi)+d_{m,n}\sin(m\theta-n_{fp}n\phi)\right]. (2b) Here, \theta is the poloidal angle, \phi is the toroidal angle, and ( M_{pol} , N_{tor} ) are the number of poloidal and toroidal modes. The cosine and sine coefficients x_{eq} =( a_{m,n},b_{m,n},c_{m,n},d_{m,n} ) define the shape of the surface and are the degrees of freedom for optimization. Quasisymmetry is highly desirable and is added in this work as an objective function. This is done through the two-term metric for quasisymmetry [ 23 ] : f_{QS}=\sum_{s_{i}}w_{i}\biggr\langle\left[\frac{1}{|\vec{B}|^{3}}\left((N-\iota M)\vec{B}\times\vec{\nabla}B\cdot\vec{\nabla}\psi-(MG+NI)\vec{B}\cdot\vec{\nabla}B\right)\right]^{2}\biggr\rangle. (3) Where \psi is the radial flux coordinate, G and I are the poloidal and toroidal currents respectively, and s_{i} correspond to the surfaces where the metric is evaluated. On top of the quasisymmetry metric, other targets were used in the optimization, namely aspect ratio A and rotational transform \iota : \displaystyle f_{\iota}=(\iota-\iota_{T})^{2}, (4a) \displaystyle f_{A}=(A-A_{T})^{2}. (4b) Where A_{T} and \iota_{T} are the target values set by the user in the optimization. The coils are modeled