Quasi-Axisymmetric Stellarator Overview

• Quasi-axisymmetric stellarators are magnetic confinement devices that mimic tokamak behavior by using a symmetry in the magnetic field strength in Boozer coordinates despite 3D geometry.
• They employ near-axis expansions and gradient-based optimizations to achieve 3D equilibria with reduced neoclassical transport and controlled bootstrap currents.
• Recent advances in coil design and hardware simplification, including permanent magnets and winding-surface methods, enhance both confinement performance and reactor viability.

A quasi-axisymmetric stellarator is a stellarator in which the magnetic-field strength is arranged so that, in Boozer coordinates, it is approximately independent of the toroidal Boozer angle even though the device is fully three-dimensional. In the standard quasisymmetric form,

$B = B(\psi, M\theta - N\zeta),$

the quasi-axisymmetric case is the special case $N=0$, so that the field-strength variation is axisymmetric-like while the geometry and coil system remain non-axisymmetric (Landreman et al., 2018, Velasco et al., 20 Mar 2026, Schuett et al., 2024). This hidden symmetry is attractive because the guiding-center dynamics depend on the magnetic field through $B$, so QA configurations can reproduce tokamak-like single-particle confinement and reduced neoclassical transport without relying on inductive plasma current as the main source of rotational transform (Landreman et al., 2018, Jorge et al., 2020). Contemporary work treats QA stellarators simultaneously as a near-axis mathematical class, an optimization target for full 3D equilibria, and an engineering problem involving coil simplification, robustness, bootstrap-current control, and reactor-scale operability (Rodriguez et al., 2022, Giuliani, 2023, Wright et al., 26 Dec 2025).

1. Definition and physical basis

In stellarator theory, quasisymmetry means that the magnetic-field strength $B = |\mathbf B|$ is independent of one Boozer angle, even though the full vector field $\mathbf B$ is not axisymmetric (Landreman et al., 2018). Near the magnetic axis, one first-order representation used for quasisymmetric fields is

$$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$

with quasi-axisymmetry corresponding to $N=0$ and quasi-helical symmetry to $N\neq 0$ (Landreman et al., 2018). Related formulations appear throughout the literature as

$B = B(\psi, M\theta_B - N\varphi_B),$

or, for QA specifically, an axisymmetric-like dependence of $B$ on a poloidal-like angle in Boozer coordinates (Schuett et al., 2024, Velasco et al., 20 Mar 2026).

The physical significance of QA follows from the structure of guiding-center motion. The cited works state that in Boozer coordinates the guiding-center Lagrangian depends on the magnetic field only through $N=0$0, so a symmetry in $N=0$1 yields tokamak-like confinement properties, reduced neoclassical transport, impurity screening, and potentially favorable flow and stability properties (Landreman et al., 2018). In the neoclassical framing of approximately QA fields, the tokamak analogue is made explicit by the observation that, in axisymmetry, $N=0$2, while in QA one seeks the same effective form in Boozer coordinates without geometric axisymmetry (Velasco et al., 20 Mar 2026).

A recurrent misconception is that QA implies an axisymmetric shape. The opposite is emphasized repeatedly: the symmetry is in the field strength, not in the real-space geometry (Landreman et al., 2018, Plunk, 2020). Another important clarification is that QA is a subtype of quasisymmetry rather than a separate concept; quasi-helical and quasi-isodynamic fields belong to neighboring parts of the same broader hidden-symmetry landscape (Yu et al., 5 May 2026, Rodriguez et al., 2022).

2. Near-axis theory and direct construction

A large part of modern QA theory derives from the Garren–Boozer near-axis expansion, which reduces the local quasisymmetry problem to ordinary differential equations along a prescribed magnetic axis (Landreman et al., 2018, Rodriguez et al., 2022). In the first-order formulation used for direct construction, nearby points are written as

$N=0$3

with

$N=0$4

(Landreman et al., 2018). The central near-axis quasisymmetry equation is the nonlinear ODE

$N=0$5

where $N=0$6 and $N=0$7 are the curvature and torsion of the magnetic axis, $N=0$8 is the on-axis rotational transform, $N=0$9 is the on-axis value of $B$0, $B$1 is the leading coefficient of the toroidal-current function, and $B$2 controls the first-order variation of $B$3 (Landreman et al., 2018). In appendix form the equation can be recast schematically as

$B$4

with $B$5, and existence and uniqueness of a periodic solution are proven (Landreman et al., 2018).

A major result is that, for any magnetic axis with nonvanishing curvature, the first-order quasisymmetric solutions form an infinite family parameterized by the axis shape $B$6, $B$7, $B$8, and $B$9 (Landreman et al., 2018). For stellarator-symmetric configurations, $B = |\mathbf B|$0 is odd and $B = |\mathbf B|$1, and if the on-axis current vanishes then $B = |\mathbf B|$2, so in many practical cases only one scalar parameter remains free once the axis shape is fixed (Landreman et al., 2018). The same literature also connects the quasisymmetry helicity $B = |\mathbf B|$3 to the topology of the magnetic axis: in the topological near-axis classification, the axis self-linking number satisfies $B = |\mathbf B|$4, with QA occupying the $B = |\mathbf B|$5 phase (Rodriguez et al., 2022).

Several papers extend the near-axis theory in complementary directions. A direct-coordinate expansion derives first- and second-order quasisymmetry conditions explicitly in physical space, verifies that the number of equations exceeds the number of parameters at third order, and independently confirms that exact global quasisymmetry is generically overdetermined beyond low order (Jorge et al., 2020). The same work also states that the near-axis expansion does not prohibit exact quasisymmetry on a single flux surface (Jorge et al., 2020). A perturbative approach instead starts from a finite-pressure axisymmetric equilibrium,

$B = |\mathbf B|$6

and constructs approximately QA equilibria as non-axisymmetric deformations of that tokamak-like base state; in this framework, external rotational transform appears at $B = |\mathbf B|$7 even though the field perturbation itself is $B = |\mathbf B|$8 (Plunk, 2020).

The near-axis construction is not merely formal. The direct-construction implementation was passed to VMEC and BOOZ_XFORM, and the resulting equilibria displayed the expected dominant Boozer harmonic; in the QA example the $B = |\mathbf B|$9 harmonic dominates, and VMEC’s rotational transform converges to the near-axis prediction as aspect ratio increases (Landreman et al., 2018). The predicted scaling of symmetry breaking is also explicit: since the construction is only first-order quasisymmetric in radius $\mathbf B$0, symmetry-breaking Fourier amplitudes are expected to scale like $\mathbf B$1, with the dominant “good” quasisymmetric mode scaling like $\mathbf B$2 and the symmetry-breaking terms smaller by an additional factor of $\mathbf B$3 (Landreman et al., 2018).

3. Design-space exploration and optimization methodologies

QA stellarators are now designed through several distinct but partially overlapping computational strategies: direct near-axis construction, fixed- or free-boundary equilibrium optimization, direct coil optimization, and combined plasma–coil single-stage algorithms (Landreman et al., 2018, Schuett et al., 2024, Giuliani et al., 2020, Jorge et al., 2024). A unifying trend is the use of quasisymmetry metrics that avoid explicit Boozer-coordinate conversion during optimization. One common form is the two-term measure

$\mathbf B$4

used with $\mathbf B$5 or equivalent QA conventions depending on the paper’s notation (Jorge et al., 2024, Gil et al., 12 Mar 2026, Buller et al., 2024). Another local criterion is the triple-product metric

$\mathbf B$6

used in DESC as a local quasisymmetry error indicator evaluated without Boozer coordinates (Salguero-Martínez et al., 4 May 2026).

Near-axis optimization has become a design-space survey tool in its own right. One approach optimized the near-axis equations to maximize the volume over which the asymptotic expansion remains accurate, with an objective

$\mathbf B$7

and identified $\mathbf B$8 as a key term because minimizing it enlarges the field scale length $\mathbf B$9 and therefore the radius of near-axis validity (Landreman, 2022). That scan produced a database of $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$0 optimized configurations and found that QA solutions exist in continuous bands, concentrated primarily at low $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$1, with the strongest and cleanest QA cases at $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$2 and some marginal $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$3 cases (Landreman, 2022). A related topological construction of the “space of quasisymmetric stellarators” organizes designs by regular axis curves and a few near-axis parameters; in that picture, existing QA devices cluster on a dominant branch in the $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$4 phase, close to a canonical choice $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$5 that extremizes the axis rotational transform and tends to regularize the surface shaping (Rodriguez et al., 2022).

Full-equilibrium optimization has extended QA far beyond the strict near-axis limit. The first numerical optimization of a class of compact QA stellarators derived analytically found improved finite-$$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$6 equilibria at aspect ratio $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$7, for field periods $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$8, with no alpha-particle losses in volume-scaled reactor comparisons and Mercier-stable configurations in some cases (Schuett et al., 2024). Another study used continuation from the “Precise QA” equilibrium to generate a family with mean rotational transform $$B(r,\theta,\phi)=B_0 + r\,\bar\eta B_0 \cos(\theta-N) + O(r^2),$$9 from $N=0$0 to $N=0$1, finding two-term quasisymmetry errors comparable to or lower than the Landreman–Paul reference over a broad interval and showing that fast-particle confinement improves with $N=0$2 until quasisymmetry degradation offsets the gain at $N=0$3 (Buller et al., 2024).

Coil-first and plasma–coil combined methods have also become central. A single-stage gradient-based method directly optimized coil shapes and currents to match the on-axis field and field gradient of a target near-axis QA configuration, solving the constrained $N=0$4-equation through forward and adjoint sensitivities (Giuliani et al., 2020). That work argued that quasi-Newton methods are crucial because the objective landscape contains many flat directions, indicating considerable latitude to simplify coils without strongly degrading the QA target (Giuliani et al., 2020). A global-to-local direct coil-design pipeline later combined a global exploration phase with three local stages—near-axis QA fitting, BoozerLS volume optimization, and BoozerExact polishing—and generated the QUASR repository of approximately 200,000 coil sets (Giuliani, 2023). The resulting database was used to study trade-offs among coil length, aspect ratio, rotational transform, and attainable QA quality, with lower aspect ratio generally making high-quality QA more difficult (Giuliani, 2023).

More recent single-stage studies show that QA can be retained while greatly simplifying the coil system. Examples include two-field-period QA stellarators with one to three coils per half field period, variants with external trim coils, helical-coil realizations, and even a single coil set capable of producing both a QA and a QH equilibrium through current reversal (Jorge et al., 2024). A stochastic single-stage framework optimized the average objective over perturbed coil ensembles, improving robustness relative to deterministic single-stage optimization; for its NCSX-like QA test case, the stochastic single-stage result achieved better fixed-boundary quasisymmetry than the standard single-stage method and comparable robustness to stochastic stage II (Gil et al., 12 Mar 2026).

4. Coil realizations, simplified hardware, and experimental embodiments

QA research has increasingly emphasized manufacturability and configuration flexibility. One line of work addresses coil-shape realism directly. In FOCUS, cubic B-splines were introduced to replace global Fourier coil representations, together with a localized outboard curvature penalty,

$N=0$5

to make the outer side of a coil straighter while leaving the inboard side freer to satisfy the magnetic target (Lonigro et al., 2021). Applied to a prototype quasi-axisymmetric reactor-sized stellarator, this produced coils whose outer side was largely straight and still reproduced the target equilibrium well in free-boundary VMEC, which the paper explicitly interprets as favorable for remote access, blanket replacement, and improved engineering maintainability (Lonigro et al., 2021).

A different simplification route replaces non-planar shaping coils with permanent magnets. For an NCSX-based QA plasma, a design using planar toroidal-field coils together with cubic rare-earth permanent magnets optimized the normal-field error

$N=0$6

and then pushed magnet strengths toward binary values using

$N=0$7

(Hammond et al., 2022). The final design used only three magnet polarization types, contained 35,436 cubic magnets per half-period, and, when extended to the full six-period device, required 212,616 magnets with total magnet volume $N=0$8 (Hammond et al., 2022). Free-boundary VMEC verification showed rotational transform within about $N=0$9 of the target and effective ripple $N\neq 0$0 near $N\neq 0$1 in the core and $N\neq 0$2 near the edge, while field-line tracing confirmed vacuum closed flux surfaces (Hammond et al., 2022).

Coil winding-surface methods provide another engineering lever. For an approximately quasisymmetric target equilibrium, filamentary coils were constrained to lie on either an axisymmetric circular toroidal winding surface or a winding surface rescaled from the plasma boundary, with optimization in SIMSOPT using the quadratic-flux objective

$N\neq 0$3

(Biu et al., 12 May 2025). The rescaled boundary surface reproduced the target more accurately, but the circular torus remained surprisingly effective, suggesting that simple axisymmetric winding surfaces may still be viable for quasisymmetric devices if some loss of fidelity is acceptable (Biu et al., 12 May 2025). In that study the optimal coil-plasma distance was about $N\neq 0$4 for both winding-surface choices in the compact five-field-period example (Biu et al., 12 May 2025).

At the device scale, the Columbia Stellarator eXperiment is a university-scale QA stellarator upgrade of CNT using two optimized non-planar interlocked HTS magnets plus two copper poloidal-field magnets, with on-axis field target $N\neq 0$5 (Schmeling et al., 8 Oct 2025). Because ReBCO is strain-sensitive under torsion and hard-way bending, the prototype program developed 3D-printed sectional aluminum coil frames, gimballed constant-tension winding, solder potting, and low-resistance joints (Schmeling et al., 8 Oct 2025). The staged P1–P3 prototypes were used to de-risk manufacturing, cooling, quench resilience, and non-planar winding, with P2 demonstrating expected fields in a non-planar high-strain geometry and P3 targeting concave geometry and higher field closer to the final CSX conditions (Schmeling et al., 8 Oct 2025). This line of work shows that QA engineering is increasingly constrained by HTS mechanics and thermal integration rather than by magnetic theory alone (Schmeling et al., 8 Oct 2025).

A still more radical hardware concept is the programmable stellarator–tokamak hybrid. This device combines tokamak-like base coils with 288 dipole-field coils on a $N\neq 0$6 toroidal-by-poloidal grid, requiring only six unique coil shapes by symmetry (Yu et al., 5 May 2026). By programming the dipole-field currents, the same hardware accesses more than 1.66 million optimized stellarator configurations spanning QA, QH, and QI, as well as tokamak-relevant 3D perturbations (Yu et al., 5 May 2026). The representative two-period QA state in that database has aspect ratio $N\neq 0$7, edge rotational transform $N\neq 0$8, and $N\neq 0$9, with nested magnetic surfaces and the lowest dipole-field current burden among the representative QA/QH/QI examples (Yu et al., 5 May 2026). This reinterprets QA not as a single custom machine but as one subset of a much larger current-programmable configuration space (Yu et al., 5 May 2026).

5. Transport, stability, and operational constraints

The primary confinement rationale for QA is the reduction of neoclassical transport and improvement of energetic-particle confinement, but the modern literature also identifies several limitations. Neoclassical performance is often characterized by effective ripple $B = B(\psi, M\theta_B - N\varphi_B),$0, a proxy for $B = B(\psi, M\theta_B - N\varphi_B),$1 transport. In the simplified four-coil free-boundary DESC study, the best finite-pressure $B = B(\psi, M\theta_B - N\varphi_B),$2 case reduced $B = B(\psi, M\theta_B - N\varphi_B),$3 by two orders of magnitude relative to poorer cases and was described as having very good Boozer-coordinate quasi-symmetry (Salguero-Martínez et al., 4 May 2026). In the programmable hybrid study, the representative QA state had $B = B(\psi, M\theta_B - N\varphi_B),$4, better than the representative QH example though not as low as the representative QI case (Yu et al., 5 May 2026). Compact QA optimization has also produced reactor-scaled equilibria with no alpha losses in volume-scaled comparisons (Schuett et al., 2024).

Rotational transform is a central QA design variable because it affects orbit width, coil burden, and stability. In the continuation study of QA families, the orbit-width argument predicts improved fast-particle confinement with increasing $B = B(\psi, M\theta_B - N\varphi_B),$5, and the calculations confirm improving confinement up to $B = B(\psi, M\theta_B - N\varphi_B),$6, beyond which worsening quasisymmetry dominates (Buller et al., 2024). The same study found that the required coil-plasma distance stayed within about $B = B(\psi, M\theta_B - N\varphi_B),$7 across the family and lay between $B = B(\psi, M\theta_B - N\varphi_B),$8 and $B = B(\psi, M\theta_B - N\varphi_B),$9 at reactor scale (Buller et al., 2024). However, higher $B$0 also increased the maximum linear ITG growth rate and shifted it toward higher $B$1, while the nonlinear ion heat flux displayed a non-monotonic dependence on $B$2; sufficiently large positive shear was destabilizing in both linear and nonlinear calculations (Buller et al., 2024).

Bootstrap current is both a feature and a liability in QA. Approximately QA stellarators are explicitly described as the stellarator analogue of the axisymmetric tokamak, retaining compactness and relative coil simplicity while avoiding the need for inductive current drive (Velasco et al., 20 Mar 2026). At the same time, QA generally implies a substantial bootstrap current, which can modify the rotational transform and complicate compatibility with island divertors (Velasco et al., 20 Mar 2026). A new strategy proposes adding a piecewise omnigenous perturbation to an approximately QA field, producing a QA-pwO geometry in which tokamak-like low-collisionality transport is preserved while the bootstrap current is reduced or even canceled through competing contour helicities (Velasco et al., 20 Mar 2026). The paper’s central claim is that this may reconcile QA-like confinement with island-divertor compatibility, a relationship that had historically been regarded as antagonistic (Velasco et al., 20 Mar 2026).

Macro-MHD optimization is not sufficient by itself. In a three-field-period, reactor-scale QA equilibrium optimized for $B$3, linear ideal MHD stability, low neoclassical transport, and self-consistent bootstrap current, high-fidelity simulations found an abrupt transition to deleterious kinetic-ballooning-mode transport at very low local $B$4, between about $B$5 and $B$6, well below the $B$7 beta limit associated with destructive macroscopic MHD instability in the earlier design study (Wright et al., 26 Dec 2025). Linear GENE simulations showed a transition from ITG-dominated behavior to KBM onset, and nonlinear calculations at $B$8 produced explosive, unsaturated transport dominated by low-$B$9 modes and strong electromagnetic heat flux (Wright et al., 26 Dec 2025). This directly challenges the assumption that favorable MHD and neoclassical metrics imply practical reactor accessibility (Wright et al., 26 Dec 2025).

QA configurations with large plasma current can also suffer violent external kink dynamics. In reduced-MHD JOREK simulations using an axisymmetric virtual-current approximation for a QA stellarator, a configuration with only a small external rotational transform, $N=0$00, remained susceptible to low-$N=0$01 external kinks when the edge safety factor was below two (Ramasamy et al., 2021). Increasing the external rotational transform reduced the violence of the initial kink and eventually stabilized the external kink in the lowest-current case, but internal modes triggered during the nonlinear phase continued to degrade confinement over much of the parameter space (Ramasamy et al., 2021). This suggests that QA operational space improves only when a significant fraction of the transform is provided externally, enough to suppress both the external kink and the induced internal activity (Ramasamy et al., 2021).

6. Reactor relevance, comparative position, and open directions

QA stellarators occupy a distinctive position among magnetic-confinement concepts. They are repeatedly described as combining tokamak-like confinement properties with stellarator steady-state operation and reduced reliance on plasma current (Landreman et al., 2018, Velasco et al., 20 Mar 2026, Schuett et al., 2024). Compared with more general stellarators, QA often offers relative coil simplicity and compactness; compared with tokamaks, it replaces inductive current drive with external 3D shaping (Velasco et al., 20 Mar 2026). This conceptual position explains the sustained interest in compact QA families, simplified coil sets, programmable hardware, and hybrid perturbative constructions that continuously connect tokamak-like and stellarator-like states (Schuett et al., 2024, Plunk, 2020, Yu et al., 5 May 2026).

The design landscape, however, is neither unique nor monotone. Near-axis studies emphasize infinite first-order solution families for a given regular axis, continuous bands of acceptable configurations, and distinct topological phases labeled by the quasisymmetry helicity (Landreman et al., 2018, Landreman, 2022, Rodriguez et al., 2022). Global coil studies likewise show many local minima and broad flat directions in the objective landscape, indicating both freedom and ambiguity in practical QA design (Giuliani et al., 2020, Giuliani, 2023). This suggests that QA is best understood not as a single optimized configuration class but as a structured high-dimensional family whose attainable properties depend strongly on aspect ratio, field-period count, transform profile, and engineering regularization (Giuliani, 2023, Buller et al., 2024).

Several points now appear settled. First, QA is not prohibited by low aspect ratio: numerical optimization has produced compact finite-$N=0$02 QA equilibria at $N=0$03 with favorable alpha confinement and Mercier-stable examples (Schuett et al., 2024). Second, QA does not require a bespoke non-planar coil architecture in every realization: simplified modular, helical, permanent-magnet, axisymmetric winding-surface, and programmable planar-coil-adjacent approaches all exist in the current literature (Jorge et al., 2024, Hammond et al., 2022, Biu et al., 12 May 2025, Yu et al., 5 May 2026). Third, exact global quasisymmetry remains mathematically restrictive, with near-axis analyses showing overdetermination beyond low order, so practical QA continues to be an exercise in controlled approximation rather than exact symmetry realization (Jorge et al., 2020, Landreman et al., 2018).

Less settled are the reactor-operability questions. The KBM accessibility study implies that low-neoclassical, macro-stable QA equilibria can still be blocked by microturbulent transitions at unexpectedly low $N=0$04 (Wright et al., 26 Dec 2025). The bootstrap-current control study suggests that the historical incompatibility between QA and island-divertor exhaust may be softened by structured departures from strict QA, specifically piecewise omnigenous perturbations (Velasco et al., 20 Mar 2026). The programmable-hybrid results further suggest that future QA research may shift from isolated hand-designed equilibria to database-scale configuration discovery, where QA is studied side by side with QH, QI, and tokamak-relevant 3D perturbations on a common hardware platform (Yu et al., 5 May 2026).

Taken together, the literature presents the quasi-axisymmetric stellarator as a rigorously defined hidden-symmetry class, a near-axis solvable family, a target of modern gradient-based and stochastic optimization, and an increasingly engineering-conscious reactor concept. Its central promise remains the same throughout: to recover much of the confinement structure associated with axisymmetry while preserving the steady-state and externally generated rotational-transform advantages of stellarators (Landreman et al., 2018, Velasco et al., 20 Mar 2026). Its central challenge is equally clear: that this hidden symmetry must be realized not only in equilibrium calculations, but also in robust coils, finite-pressure operation, bootstrap-current control, turbulence-limited regimes, and experimentally buildable hardware (Lonigro et al., 2021, Wright et al., 26 Dec 2025, Schmeling et al., 8 Oct 2025).