betaFeedback

WorksPhysics and Astronomy

Non-axisymmetric MHD equilibria with an arbitrary number of field periods

Abstract

Several closed-form non-axisymmetric magnetic fields with nested toroidal flux surfaces have recently been discovered as exact benchmarks for verifying three-dimensional magnetohydrodynamic equilibrium codes. However, all of these solutions have at most two field periods (N≤2N \le 2), i.e., identical segments that repeat toroidally. Here, we derive closed-form non-axisymmetric magnetohydrodynamic equilibria for an arbitrary integer number of field periods N≥1N \geq 1. The equilibria have non-constant pressure and closed magnetic field lines with either zero rotational transform (ι=0ι= 0) for N≥1N \geq 1 or integer rotational transform (ι=±Nι= \pm N) for N≥3N \ge 3. For the reflection-symmetric ι=0ι= 0 equilibria, we also construct a matched exterior vacuum field with continuous magnetic field across the plasma boundary, providing a benchmark for free-boundary equilibrium codes.

Introduction

In Cartesian coordinates x⃗:=(x,y,z)\vec{x} := (x,y,z), the magnetic field B⃗(x⃗):=(Bx,By,Bz)\vec{B}(\vec{x}) := (B_x,B_y,B_z), current density J⃗(x⃗):=∇×B⃗\vec{J}(\vec{x}) := \nabla\times\vec{B}, and scalar plasma pressure p(x⃗)p(\vec{x}) of a static ideal magnetohydrodynamic (MHD) plasma satisfy

J⃗×B⃗=∇p,∇⋅B⃗=0,B⃗⋅∇ψ=0,(1)\vec{J} \times\vec{B} = \nabla p,\qquad\nabla\cdot\vec{B} = 0,\qquad\vec{B} \cdot\nabla\psi= 0, \tag*{(1)}

where the pressure p(ψ)p(\psi) depends on the flux function ψ(x⃗)∈[0,δ]\psi(\vec{x}) \in[0,\delta] whose level sets form compact nested tori around a magnetic axis at ψ=0\psi= 0 up to a boundary surface ψ=δ\psi= \delta. We also use cylindrical coordinates (R,ϕ,z)(R,\phi,z) defined by x=Rcos⁡(ϕ)x = R\cos(\phi) and y=Rsin⁡(ϕ)y = R\sin(\phi).

Assuming axisymmetry (∂ϕ=0\partial_\phi= 0), the static ideal MHD equations (1) simplify to the two-dimensional Grad–Shafranov equation, which admits many known exact solutions [1, 2, 3, 4]. Without axisymmetry, resonant Pfirsch–Schlüter currents on rational surfaces [5, 6] prompted Grad [7, 8] to conjecture that smooth toroidal equilibria with non-constant pressure do not exist. Nevertheless, Lortz [9] proved more than five decades ago the existence of reflection-symmetric equilibria with zero rotational transform (ι=0\iota= 0) via a fixed-point iteration, where ι\iota is the ratio of poloidal to toroidal turns of a magnetic field line (positive or negative depending on the winding direction). More recently, Gómez-Serrano et al. [10] established the existence of smooth non-axisymmetric toroidal equilibria with discrete rotational symmetry and infinite rotational transform (ι=∞\iota= \infty). However, neither proof provides explicit closed-form solutions.

Several closed-form three-dimensional toroidal magnetic fields with nested flux surfaces were discovered very recently [11, 12, 13, 14]. In the vacuum limit of J⃗=0⃗\vec{J} = \vec{0} and ∇p=0⃗\nabla p = \vec{0}, Sengupta et al. [13] constructed closed-form magnetic fields with constant (rational or irrational) or sheared rotational transform using Jacobi elliptic functions, and López [14] constructed a closed-form magnetic field with ι=−1\iota= -1 via a cubic Weierstrass polynomial. For non-constant pressure, Landreman [11] derived two closed-form stellarator-symmetric MHD equilibria, one with ι=2\iota= 2 and one with sheared rotational transform ι′(ψ)≠0\iota'(\psi) \ne0, which Issan et al. [12] generalized to equilibria without stellarator symmetry [15].

A non-axisymmetric toroidal equilibrium has NN field periods if it satisfies the following discrete rotational symmetry

ψ(R,ϕ+2π/N,z)=ψ(R,ϕ,z),Bj(R,ϕ+2π/N,z)=Bj(R,ϕ,z)for j∈{R,ϕ,z}.(2)\psi(R,\phi+ 2\pi/N,z) = \psi(R,\phi,z), \qquad B_j(R,\phi+ 2\pi/N,z) = B_j(R,\phi,z) \quad\text{for } j \in\{R,\phi,z\}. \tag*{(2)}

To the best of our knowledge, all recent closed-form solutions [11, 12, 13, 14] are restricted to at most two field periods (N≤2N \le2). In practice, most stellarators are designed with N≥3N \ge3 field periods, including NCSX (N=3N = 3) [16], HSX and TJ-II (N=4N = 4) [17, 18], Wendelstein 7-X (N=5N = 5) [19, 20], and LHD (N=10N = 10) [21]. Extending analytical solutions to N≥3N \ge3 therefore broadens the variety of exact benchmarks available for verifying three-dimensional equilibrium codes such as VMEC [22], DESC [23], GVEC [24], and SPEC [25].

In this paper, we construct closed-form non-axisymmetric MHD equilibria for an arbitrary integer number of field periods N≥1N \ge1. All solutions originate from a coordinate transformation that maps field-aligned coordinates to Cartesian coordinates via a three-harmonic Fourier expansion in the field-aligned toroidal angle. This coordinate ansatz reduces the three-dimensional nonlinear MHD partial differential equations to a sequence of separable first-order linear ordinary differential equations and algebraic relations. Solving this system of equations yields three distinct families of equilibria, each featuring nested toroidal flux surfaces, non-constant plasma pressure, stellarator symmetry, and closed magnetic field lines. The first family has zero rotational transform (ι=0\iota= 0) for any number of field periods N≥1N \ge1 and provides an explicit closed-form realization of the non-axisymmetric equilibria proven to exist by Lortz [9]. The second family features an ι=0\iota= 0 branch for N≥2N \ge2 and an integer rotational transform ι=±N\iota= \pm N branch for N≥3N \ge3, whereas the third family yields equilibria with ι=±N\iota= \pm N for N≥3N \ge3. The ι=±N\iota= \pm N branch of the second family is restricted to narrow cross-sections, while the third family has much broader ones with a larger plasma volume and beta. Table 1 compares the equilibria derived in this work with recent closed-form non-axisymmetric solutions [11, 12, 13, 14].

PropertyLandreman [11]Issan et al. [12]Sengupta et al. [13]López [14]Present Work
Physical regime∇p≠0⃗, J⃗≠0⃗\nabla p \ne\vec{0},\ \vec{J} \ne\vec{0}∇p≠0⃗, J⃗≠0⃗\nabla p \ne\vec{0},\ \vec{J} \ne\vec{0}∇p=0⃗, J⃗=0⃗\nabla p = \vec{0},\ \vec{J} = \vec{0}∇p=0⃗, J⃗=0⃗\nabla p = \vec{0},\ \vec{J} = \vec{0}∇p≠0⃗, J⃗≠0⃗\nabla p \ne\vec{0},\ \vec{J} \ne\vec{0}
Analytical formElementaryElementaryJacobi ellipticElementaryElementary
Field periods NNN=2N = 2N=2N = 2N=2N = 2N=1N = 1N≥1N \ge1
Rotational transform ι(ψ)\iota(\psi)ι=2\iota= 2 or sheared ι′(ψ)≠0\iota'(\psi) \ne0ι=2\iota= 2 or sheared ι′(ψ)≠0\iota'(\psi) \ne0Constant or sheared ι′(ψ)≠0\iota'(\psi) \ne0ι=−1\iota= -1ι=0\iota= 0 (N≥1N \ge1) or ι=±N\iota= \pm N (N≥3N \ge3)
Stellarator symmetry✓✗/✓✗/✓✗✓

Table 1. Comparison of closed-form non-axisymmetric toroidal solutions.

This paper is organized as follows. § 2 presents the coordinate map ansatz and the three families of equilibria. § 3 verifies the nested toroidal flux surfaces, Hamada condition, and discrete symmetries. § 4 characterizes the magnetic axis, enclosed plasma volume, volume-averaged plasma beta, and ideal-MHD interchange stability. § 5 constructs a matched exterior vacuum field for ι=0\iota= 0 equilibria, providing a benchmark for free-boundary equilibrium codes. § 6 summarizes the results and discusses future work, and Appendix § 8 presents a detailed derivation of the equilibria.

Equilibria with an arbitrary number of field periods NN

We begin by introducing field-aligned curvilinear coordinates (ρ,ζ,η)(\rho,\zeta,\eta), consisting of a toroidal angle ζ∈[0,2π)\zeta\in[0,2\pi) along the magnetic field lines together with radial and vertical field-line labels ρ>0\rho> 0 and η∈R\eta\in\mathbb{R}, respectively. The coordinate ansatz relating the field-aligned coordinates (ρ,ζ,η)(\rho,\zeta,\eta) to the Cartesian coordinates (x,y,z)(x,y,z) is

x+iy=C1(ρ)eim1ζ+(C2(ρ)+iηC4(ρ))eim2ζ−C3(ρ)eim3ζ,z=η+C5(ρ)sin⁡(Nζ),(3)x + iy = C_1(\rho)e^{im_1\zeta} + \left(C_2(\rho) + i\eta C_4(\rho)\right)e^{im_2\zeta} - C_3(\rho)e^{im_3\zeta}, \qquad z = \eta+ C_5(\rho)\sin(N\zeta), \tag*{(3)}

where the toroidal mode numbers (m1,m2,m3)(m_1,m_2,m_3) are

m1:=1+(k−1)N,m2:=1+kN,m3:=1+(k+1)N,for k∈{1,0,−1}.(4)m_1 := 1 + (k-1)N,\qquad m_2 := 1 + kN,\qquad m_3 := 1 + (k+1)N,\qquad\text{for } k \in\{1,0,-1\}. \tag*{(4)}

For a length-scale parameter ρ0>0\rho_0 > 0 and magnetic axis location (ρ,η)=(ρa,0)(\rho,\eta) = (\rho_a,0), the auxiliary functions in Eq. (3) are

C1(ρ):=sign⁡(m1)ρ0m12(ρρ0)m1,C2(ρ):=C2(ρa)(ρaρ)ND−(ρa)D−(ρ),C3(ρ):=ρ0m32(ρ0ρ)m3,C4(ρ):=N2C5(ρ)D+(ρ),C5(ρ):=2ρ0N∣m1∣m3(ρ0ρ)N.(5)\begin{aligned} C_1(\rho) &:= \frac{\operatorname{sign}(m_1)\rho_0}{m_1^2}\left(\frac{\rho}{\rho_0}\right)^{m_1}, \qquad C_2(\rho) := C_2(\rho_a)\left(\frac{\rho_a}{\rho}\right)^N\frac{D_-(\rho_a)}{D_-(\rho)},\\ C_3(\rho) &:= \frac{\rho_0}{m_3^2}\left(\frac{\rho_0}{\rho}\right)^{m_3}, \qquad C_4(\rho) := \frac{N^2 C_5(\rho)}{D_+(\rho)},\\ C_5(\rho) &:= \frac{2\rho_0}{N\sqrt{\lvert m_1\rvert m_3}}\left(\frac{\rho_0}{\rho}\right)^N. \tag*{(5)} \end{aligned}

where

D±(ρ):=m1(m1+m2)C1(ρ)±m3(m3+m2)C3(ρ),D_{\pm}(\rho) := m_1(m_1+m_2)C_1(\rho) \pm m_3(m_3+m_2)C_3(\rho),
C2(ρa):=s(m1C1(ρa)+m3C3(ρa))(m12C1(ρa)−m32C3(ρa))D−(ρa)m22D+(ρa),s∈{−1,1}.(6)C_2(\rho_a) := s\sqrt{\frac{(m_1C_1(\rho_a)+m_3C_3(\rho_a))(m_1^2C_1(\rho_a)-m_3^2C_3(\rho_a))D_-(\rho_a)}{m_2^2D_+(\rho_a)}}, \qquad s \in\{-1,1\}. \tag*{(6)}

Aligning the magnetic field with the ζ\zeta-coordinate curves via B⃗:=∂ζx⃗\vec{B} := \partial_{\zeta}\vec{x} results in

Bx+iBy=im1C1(ρ)eim1ζ+im2(C2(ρ)+iηC4(ρ))eim2ζ−im3C3(ρ)eim3ζ,Bz=NC5(ρ)cos⁡(Nζ).(7)\begin{aligned} B_x+iB_y &= im_1C_1(\rho)e^{im_1\zeta}+im_2(C_2(\rho)+i\eta C_4(\rho))e^{im_2\zeta}-im_3C_3(\rho)e^{im_3\zeta},\\ B_z &= NC_5(\rho)\cos(N\zeta). \tag*{(7)} \end{aligned}

The flux function ψ(ρ,η)\psi(\rho,\eta) and plasma pressure p(ψ)p(\psi) with on-axis pressure p0>0p_0 > 0 are

ψ(ρ,η):=H(ρ,η)−H(ρa,0)2,p(ψ):=p0−2ψ,H(ρ,η):=(m1C1(ρ)+m3C3(ρ))2+m22(C2(ρ)2+η2C4(ρ)2).(8)\begin{aligned} \psi(\rho,\eta) &:= \frac{H(\rho,\eta)-H(\rho_a,0)}{2},\\ p(\psi) &:= p_0-2\psi,\\ H(\rho,\eta) &:= (m_1C_1(\rho)+m_3C_3(\rho))^2+m_2^2(C_2(\rho)^2+\eta^2C_4(\rho)^2). \tag*{(8)} \end{aligned}

Appendix § 8 derives Eqs. (3)–(8) from the static ideal MHD equations (1). Briefly, the ansatz (3) arises from perturbing the axisymmetric map of x+iy=ρeiζx+iy=\rho e^{i\zeta} and z=ηz=\eta with an NN-periodic vertical oscillation sin⁡(Nζ)\sin(N\zeta) in zz, which requires three harmonics (m1,m2,m3)(m_1,m_2,m_3) spaced by NN in x+iyx+iy to satisfy the force-balance and divergence-free equations (1).

The coordinate Jacobian determinant J\mathcal{J} of Eq. (3) is independent of ζ\zeta, i.e.

J(ρ,η):=det⁡(∂x⃗∂(ρ,ζ,η))=D−(ρ)ρD+(ρ)((m1C1(ρ)+m3C3(ρ))2−m22η2C4(ρ)2)−D+(ρ)ρD−(ρ)m22C2(ρ)2.(9)\begin{aligned} \mathcal{J}(\rho,\eta) &:= \det\left(\frac{\partial\vec{x}}{\partial(\rho,\zeta,\eta)}\right)\\ &= \frac{D_-(\rho)}{\rho D_+(\rho)}\left((m_1C_1(\rho)+m_3C_3(\rho))^2-m_2^2\eta^2C_4(\rho)^2\right)-\frac{D_+(\rho)}{\rho D_-(\rho)}m_2^2C_2(\rho)^2. \tag*{(9)} \end{aligned}

which ensures that ∇⋅B⃗=J−1∂ζJ=0\nabla\cdot\vec{B} = \mathcal{J}^{-1}\partial_{\zeta}\mathcal{J} = 0. For the coordinate map in Eq. (3) to remain smooth and invertible on the toroidal domain, the denominators D±(ρ)D_{\pm}(\rho) in Eq. (6) and the Jacobian determinant J(ρ,η)\mathcal{J}(\rho,\eta) in Eq. (9) must remain non-zero.

We categorize the equilibria into three families based on k∈{1,0,−1}k \in\{1,0,-1\} in Eq. (4). In all numerical examples in this paper, we set ρ0=1\rho_0 = 1 and p0=2δp_0 = 2\delta for simplicity, so that the pressure vanishes on the boundary flux surface, p(δ)=0p(\delta) = 0. The remaining parameters are the number of field periods NN, the sign s∈{−1,1}s \in\{-1,1\} in Eq. (6), the magnetic axis parameter ρa>0\rho_a > 0, and the boundary flux δ>0\delta> 0, which are restricted by D±(ρ)D_{\pm}(\rho) and J(ρ,η)\mathcal{J}(\rho,\eta) remaining non-zero. Below, we state the conditions on ρa\rho_a for each family and summarize their properties in Table 2.

PropertyFamily 1Family 2Family 3
Harmonic shift index kkk=1k = 1k=0k = 0k=−1k = -1
Mode numbers (m1,m2,m3)(m_1,m_2,m_3)(1,N+1,2N+1)(1,N+1,2N+1)(1−N,1,1+N)(1-N,1,1+N)(1−2N,1−N,1)(1-2N,1-N,1)
Field periods NNN≥1N \ge1N≥2N \ge2N≥3N \ge3
Rotational transform ι\iotaι=0\iota= 0ι=0\iota= 0 (N≥2N \ge2), ι=±N\iota= \pm N (N≥3N \ge3)ι=±N\iota= \pm N

Table 2. Summary of the three families of equilibria generated by Eqs. (3)–(8).

Family 1. Setting k=1k = 1 results in (m1,m2,m3)=(1,N+1,2N+1)(m_1,m_2,m_3) = (1,N+1,2N+1), and choosing ρa≥ρ0\rho_a \ge\rho_0 yields nested toroidal flux surfaces with ι=0\iota= 0 for any number of field periods N≥1N \ge1. Figure 1 shows the boundary flux surface ψ=δ=0.02\psi= \delta= 0.02 shaded by the magnetic field amplitude ∣B⃗∣|\vec{B}| along with the magnetic axis (black curve) and a magnetic field line on the boundary surface (red curve) for various numbers of field periods NN.

Figure 1. Family 1 outer toroidal flux surfaces ψ=δ=0.02\psi= \delta= 0.02 with ρa=ρ0\rho_a = \rho_0 illustrating the magnetic amplitude ∣B⃗∣|\vec{B}| alongside the magnetic axis (black curve) and a magnetic field line on the boundary surface (red curve) for various numbers of field periods NN.

Family 2. Setting k=0k = 0 results in (m1,m2,m3)=(1−N,1,1+N)(m_1,m_2,m_3) = (1-N,1,1+N) and requires N≥2N \ge2. For N=2N = 2, any ρa>0\rho_a > 0 yields nested toroidal flux surfaces with ι=0\iota= 0. For N≥3N \ge3, choosing 0<ρa<ρmid0 < \rho_a < \rho_{\mathrm{mid}} gives ι=0\iota= 0, whereas choosing ρa>ρmid\rho_a > \rho_{\mathrm{mid}} with D+(ρa)>0D_{+}(\rho_a) > 0 and ∂ρ2H(ρa,0)>0\partial_{\rho}^{2}H(\rho_a,0) > 0 gives ι=−sN=±N\iota= -sN = \pm N, where

ρmid:=ρ0((N+2)(N−1)2(N−2)(N+1)2)1/4.\rho_{\mathrm{mid}} := \rho_0 \left(\frac{(N+2)(N-1)^2}{(N-2)(N+1)^2}\right)^{1/4}.

Family 3. For k=−1k = -1 and N≥3N \ge3, the mode numbers are (m1,m2,m3)=(1−2N,1−N,1)(m_1,m_2,m_3) = (1-2N,1-N,1), and choosing

0<ρa<ρ0(3N−2(2N−1)(N−2))1/(2N−2)0 < \rho_a < \rho_0 \left(\frac{3N-2}{(2N-1)(N-2)}\right)^{1/(2N-2)}

yields nested toroidal flux surfaces with ι=−sN=±N\iota= -sN = \pm N over a much wider range of δ\delta than the ι=±N\iota= \pm N branch of Family 2.

Figure 2 shows the boundary flux surfaces ψ=δ\psi= \delta shaded by ∣B⃗∣|\vec{B}| together with the magnetic axis (black curve) and a boundary field line (red curve) for Family 2 (δ=0.005\delta= 0.005) and Family 3 (δ=0.05\delta= 0.05) at N=ι=3N = \iota= 3 and N=ι=5N = \iota= 5. Although both equilibria have ι=N\iota= N, they represent distinct equilibria due to their different frequencies (m1,m2,m3)(m_1,m_2,m_3) in Eq. (4), with highest harmonic magnitude N+1N+1 in Family 2 versus 2N−12N-1 in Family 3. In addition, nested flux surfaces break down at a much smaller boundary flux in Family 2 (δ≈0.006\delta\approx0.006 at N=5N = 5) than in Family 3 (δ≈0.05\delta\approx0.05 at N=5N = 5).

Figure 2. Same as Figure 1 with ι=N\iota= N for Family 2 with δ=0.005\delta= 0.005 at (a) N=3N = 3 (ρa=1.12\rho_a = 1.12) and (b) N=5N = 5 (ρa=1.04\rho_a = 1.04) and for Family 3 with δ=0.05\delta= 0.05 at (c) N=3N = 3 (ρa=0.97\rho_a = 0.97) and (d) N=5N = 5 (ρa=0.88\rho_a = 0.88).

Equilibria symmetries and verification

We first verify that the equilibria defined in Eqs. (3)–(8) form smooth nested toroidal flux surfaces in § 3.1, and then verify the Hamada condition in § 3.2. We then demonstrate the discrete NN-fold rotational symmetry in § 3.3, reflection symmetry in § 3.4, and stellarator symmetry in § 3.5.

Nested toroidal flux surfaces

To verify that the level sets 0<ψ≤δ0 < \psi\le\delta form nested toroidal flux surfaces around the magnetic axis (ρ,η)=(ρa,0)(\rho, \eta) = (\rho_a, 0) in the parameter regimes of § 2, we introduce the auxiliary coordinate

r(ρ):=sign⁡(ρ−ρa)H(ρ,0)−H(ρa,0).r(\rho) := \operatorname{sign}(\rho-\rho_a)\sqrt{H(\rho,0)-H(\rho_a,0)}.

The flux function ψ\psi in Eq. (8) expressed in terms of r(ρ)r(\rho) can be written as

ψ(ρ,η)=r(ρ)2+m22η2C4(ρ)22.\psi(\rho,\eta)=\frac{r(\rho)^2+m_2^2\eta^2 C_4(\rho)^2}{2}.

Then, introducing a field-line angle θ∈[0,2π)\theta\in[0,2\pi) via

r(ρ)=2ψcos⁡(θ),m2ηC4(ρ)=2ψsin⁡(θ),r(\rho)=\sqrt{2\psi}\cos(\theta),\qquad m_2\eta C_4(\rho)=\sqrt{2\psi}\sin(\theta),

maps each surface of constant ψ\psi to a circle of radius 2ψ\sqrt{2\psi} in the (r,m2ηC4)(r,m_2\eta C_4) plane, so the flux surfaces are nested wherever the coordinate map in Eq. (3) remains smooth and invertible.

Hamada condition

When magnetic field lines are closed and the pressure is non-constant, keeping the parallel current single-valued requires ∮∣B∣−1 dℓ\oint|\mathbf{B}|^{-1}\,\mathrm{d}\ell to be constant on each flux surface, known as the Hamada condition [9, 26, 27]. Because Eq. (3) traces each closed field line once over ζ∈[0,2π)\zeta\in[0, 2\pi) with arc-length element dℓ=∣∂x/∂ζ∣ dζ=∣B∣ dζ\mathrm{d}\ell= |\partial\mathbf{x}/\partial\zeta|\,\mathrm{d}\zeta= |\mathbf{B}|\,\mathrm{d}\zeta, all three families of equilibria satisfy

∮dℓ∣B∣=∫02πdζ=2π,\oint\frac{\mathrm{d}\ell}{|\mathbf{B}|} = \int_{0}^{2\pi} \mathrm{d}\zeta= 2\pi,

which is constant across all field lines.

Discrete NN-fold rotational symmetry

An equilibrium has discrete NN-fold rotational symmetry if its flux surfaces and magnetic field are invariant under a toroidal rotation by one field period, i.e. (R,ϕ,z)↦(R,ϕ+2π/N,z)(R,\phi,z) \mapsto(R,\phi+2\pi/N,z), as defined in Eq. (2). To verify this symmetry in field-aligned coordinates, we shift the toroidal angle by one field period (ρ,ζ,η)↦(ρ,ζ+2π/N,η)(\rho,\zeta,\eta) \mapsto(\rho,\zeta+2\pi/N,\eta). From Eq. (4), each mode number mjm_j for j∈{1,2,3}j \in\{1,2,3\} equals 1 plus an integer multiple of NN, so

eimj(ζ+2π/N)=ei2π/Neimjζ.e^{\mathrm{i}m_j(\zeta+2\pi/N)} = e^{\mathrm{i}2\pi/N}e^{\mathrm{i}m_j\zeta}.

Substituting (ρ,ζ+2π/N,η)(\rho, \zeta+ 2\pi/N, \eta) into Eq. (3) therefore rotates (x,y)(x,y) by 2π/N2\pi/N and leaves zz unchanged:

x(ρ,ζ+2π/N,η)=Rcos⁡(ϕ+2π/N),y(ρ,ζ+2π/N,η)=Rsin⁡(ϕ+2π/N),z(ρ,ζ+2π/N,η)=z(ρ,ζ,η),(10)\begin{aligned} x(\rho,\zeta+2\pi/N,\eta)&=R\cos(\phi+2\pi/N),\\ y(\rho,\zeta+2\pi/N,\eta)&=R\sin(\phi+2\pi/N),\\ z(\rho,\zeta+2\pi/N,\eta)&=z(\rho,\zeta,\eta), \tag*{(10)} \end{aligned}

which confirms that (ρ,ζ,η)↦(ρ,ζ+2π/N,η)(\rho,\zeta,\eta)\mapsto(\rho,\zeta+2\pi/N,\eta) maps (R,ϕ,z)↦(R,ϕ+2π/N,z)(R,\phi,z)\mapsto(R,\phi+2\pi/N,z). Since B⃗=∂x⃗/∂ζ\vec{B}=\partial\vec{x}/\partial\zeta, differentiating Eq. (10) with respect to ζ\zeta rotates (Bx,By)(B_x,B_y) by the same angle 2π/N2\pi/N and leaves BzB_z unchanged, so the cylindrical components BRB_R, BϕB_\phi, and BzB_z are invariant under (R,ϕ,z)↦(R,ϕ+2π/N,z)(R,\phi,z)\mapsto(R,\phi+2\pi/N,z). Together with the ζ\zeta-independence of ψ(ρ,η)\psi(\rho,\eta), this confirms Eq. (2).

Reflection symmetry

In proving the existence of non-axisymmetric ι=0\iota=0 equilibria, [9] considered configurations that are symmetric across a poloidal plane ϕ=ϕ∗\phi=\phi_*. Under the reflection (R,ϕ,z)↦(R,2ϕ∗−ϕ,z)(R,\phi,z)\mapsto(R,2\phi_*-\phi,z), reflection symmetry requires

ψ(R,2ϕ∗−ϕ,z)=ψ(R,ϕ,z),BR(R,2ϕ∗−ϕ,z)=−BR(R,ϕ,z),Bϕ(R,2ϕ∗−ϕ,z)=Bϕ(R,ϕ,z),Bz(R,2ϕ∗−ϕ,z)=−Bz(R,ϕ,z).(11)\begin{aligned} \psi(R,2\phi_*-\phi,z)&=\psi(R,\phi,z),& B_R(R,2\phi_*-\phi,z)&=-B_R(R,\phi,z),\\ B_\phi(R,2\phi_*-\phi,z)&=B_\phi(R,\phi,z),& B_z(R,2\phi_*-\phi,z)&=-B_z(R,\phi,z). \tag*{(11)} \end{aligned}

Setting ϕ=ϕ∗\phi=\phi_* in Eq. (11) gives BR=Bz=0B_R=B_z=0, so magnetic field lines intersect the plane ϕ=ϕ∗\phi=\phi_* perpendicularly.

In Family 1 with ρa=ρ0\rho_a=\rho_0, we have C2(ρ)=0C_2(\rho)=0 from Eq. (6), so substituting (ρ,ζ,η)↦(ρ,π/N−ζ,η)(\rho,\zeta,\eta)\mapsto(\rho,\pi/N-\zeta,\eta) into Eq. (3) results in

x(ρ,π/N−ζ,η)=Rcos⁡(π/N−ϕ),y(ρ,π/N−ζ,η)=Rsin⁡(π/N−ϕ),z(ρ,π/N−ζ,η)=z(ρ,ζ,η),(12)\begin{aligned} x(\rho,\pi/N-\zeta,\eta)&=R\cos(\pi/N-\phi),\\ y(\rho,\pi/N-\zeta,\eta)&=R\sin(\pi/N-\phi),\\ z(\rho,\pi/N-\zeta,\eta)&=z(\rho,\zeta,\eta), \tag*{(12)} \end{aligned}

which confirms that (ρ,ζ,η)↦(ρ,π/N−ζ,η)(\rho,\zeta,\eta)\mapsto(\rho,\pi/N-\zeta,\eta) maps (R,ϕ,z)↦(R,π/N−ϕ,z)(R,\phi,z)\mapsto(R,\pi/N-\phi,z). Since B⃗=∂x⃗/∂ζ\vec{B}=\partial\vec{x}/\partial\zeta, differentiating Eq. (12) with respect to ζ\zeta and projecting onto the cylindrical basis, together with the ζ\zeta-independence of ψ(ρ,η)\psi(\rho,\eta), confirms Eq. (11) across ϕ∗=π/2N\phi_*=\pi/2N.

Family 1 with ρa=ρ0\rho_a=\rho_0 therefore provides an explicit closed-form realization of the reflection-symmetric ι=0\iota=0 equilibrium class of [9]. By contrast, Family 1 with ρa>ρ0\rho_a>\rho_0, Family 2, and Family 3 do not satisfy reflection symmetry (11) across any poloidal plane.

Stellarator symmetry

An equilibrium has stellarator symmetry [15] if it is invariant under a π\pi-rotation about the xx-axis (ϕ=0\phi=0). This rotation maps (R,ϕ,z)↦(R,−ϕ,−z)(R,\phi,z)\mapsto(R,-\phi,-z) and requires

ψ(R,−ϕ,−z)=ψ(R,ϕ,z),BR(R,−ϕ,−z)=−BR(R,ϕ,z),Bϕ(R,−ϕ,−z)=Bϕ(R,ϕ,z),Bz(R,−ϕ,−z)=Bz(R,ϕ,z).(13)\begin{aligned} \psi(R,-\phi,-z)&=\psi(R,\phi,z),& B_R(R,-\phi,-z)&=-B_R(R,\phi,z),\\ B_\phi(R,-\phi,-z)&=B_\phi(R,\phi,z),& B_z(R,-\phi,-z)&=B_z(R,\phi,z). \tag*{(13)} \end{aligned}

To verify this symmetry in field-aligned coordinates for all three families of equilibria, we substitute (ρ,ζ,η)↦(ρ,−ζ,−η)(\rho,\zeta,\eta)\mapsto(\rho,-\zeta,-\eta) into Eq. (3), i.e.

x(ρ,−ζ,−η)=Rcos⁡(−ϕ),y(ρ,−ζ,−η)=Rsin⁡(−ϕ),z(ρ,−ζ,−η)=−z(ρ,ζ,η),(14)\begin{aligned} x(\rho,-\zeta,-\eta)&=R\cos(-\phi),\\ y(\rho,-\zeta,-\eta)&=R\sin(-\phi),\\ z(\rho,-\zeta,-\eta)&=-z(\rho,\zeta,\eta), \tag*{(14)} \end{aligned}

which confirms that (ρ,ζ,η)↦(ρ,−ζ,−η)(\rho,\zeta,\eta)\mapsto(\rho,-\zeta,-\eta) maps (R,ϕ,z)↦(R,−ϕ,−z)(R,\phi,z)\mapsto(R,-\phi,-z). Since B⃗=∂x⃗/∂ζ\vec{B}=\partial\vec{x}/\partial\zeta, differentiating Eq. (14) with respect to ζ\zeta gives

B⃗(ρ,−ζ,−η)=(−Bx(ρ,ζ,η),By(ρ,ζ,η),Bz(ρ,ζ,η)).\vec{B}(\rho,-\zeta,-\eta)=(-B_x(\rho,\zeta,\eta),B_y(\rho,\zeta,\eta),B_z(\rho,\zeta,\eta)).

Projecting B⃗\vec{B} onto the cylindrical basis, together with ψ(ρ,−η)=ψ(ρ,η)\psi(\rho,-\eta)=\psi(\rho,\eta), confirms Eq. (13).

Figure 3 shows the poloidal cross-sections of the flux surfaces across four cylindrical azimuthal slices ϕ∈{0,π/2N,π/N,3π/2N}\phi\in\{0,\pi/2N,\pi/N,3\pi/2N\} at N=3N=3 across all three families of equilibria. By stellarator symmetry (13) and discrete NN-fold rotational symmetry (2), the cross-sections at ϕ=0\phi=0 and ϕ=π/N\phi=\pi/N are each symmetric across z=0z=0, while the slices at ϕ=π/2N\phi=\pi/2N and ϕ=3π/2N\phi=3\pi/2N are vertical reflections of one another across z=0z=0. In Figure 3a (C2(ρ)=0C_2(\rho)=0), reflection symmetry (11) across ϕ∗=π/2N\phi_*=\pi/2N makes the ϕ=0\phi=0 and ϕ=π/N\phi=\pi/N cross-sections coincide, whereas in Figure 3b–Figure 3d (C2(ρa)≠0C_2(\rho_a)\ne0) the ϕ=0\phi=0 and ϕ=π/N\phi=\pi/N cross-sections are distinct. In Figure 3c, the cross-sections on the ι=N\iota=N branch of Family 2 are much narrower because ∣J∣\lvert J\rvert is small near the magnetic axis, which compresses the flux surfaces and restricts δ\delta to small values where the flux surfaces remain nested.

Figure 3. Poloidal cross-sections at N=3N=3 for (a) Family 1 (ρa=1\rho_a=1, δ=0.02\delta=0.02, ι=0\iota=0), (b) Family 2 (ρa=0.75\rho_a=0.75, δ=0.32\delta=0.32, ι=0\iota=0), (c) Family 2 (ρa=1.12\rho_a=1.12, δ=0.02\delta=0.02, ι=3\iota=3), and (d) Family 3 (ρa=0.97\rho_a=0.97, δ=0.05\delta=0.05, ι=3\iota=3).

Equilibria geometry and physical properties

We now characterize the magnetic axis geometry in § 4.1, the enclosed plasma volume V(δ)V(\delta) and volume-averaged plasma beta βV(δ)\beta_V(\delta) in Section 4.2, and ideal-MHD interchange stability in § 4.3.

Magnetic axis

Evaluating Eq. (3) at (ρ,η)=(ρa,0)(\rho,\eta)=(\rho_a,0) gives the magnetic axis x⃗axis(ζ)\vec{x}_{\mathrm{axis}}(\zeta), i.e.

xaxis(ζ)+iyaxis(ζ)=C1(ρa)eim1ζ+C2(ρa)eim2ζ−C3(ρa)eim3ζ,zaxis(ζ)=C5(ρa)sin⁡(Nζ).(15)x_{\mathrm{axis}}(\zeta)+iy_{\mathrm{axis}}(\zeta)=C_1(\rho_a)e^{im_1\zeta}+C_2(\rho_a)e^{im_2\zeta}-C_3(\rho_a)e^{im_3\zeta}, \qquad z_{\mathrm{axis}}(\zeta)=C_5(\rho_a)\sin(N\zeta). \tag*{(15)}

The squared magnetic axis cylindrical radius is

Raxis(ζ)2:=∣xaxis(ζ)+iyaxis(ζ)∣2=C1(ρa)2+C2(ρa)2+C3(ρa)2+2C2(ρa)(C1(ρa)−C3(ρa))cos⁡(Nζ)−2C1(ρa)C3(ρa)cos⁡(2Nζ).(16)R_{\mathrm{axis}}(\zeta)^2:=\left|x_{\mathrm{axis}}(\zeta)+iy_{\mathrm{axis}}(\zeta)\right|^2=C_1(\rho_a)^2+C_2(\rho_a)^2+C_3(\rho_a)^2+2C_2(\rho_a)\left(C_1(\rho_a)-C_3(\rho_a)\right)\cos(N\zeta)-2C_1(\rho_a)C_3(\rho_a)\cos(2N\zeta). \tag*{(16)}

In Eq. (15), the mj=1m_j=1 harmonic represents the axisymmetric part, while the other two harmonics produce the cos⁡(Nζ)\cos(N\zeta) and cos⁡(2Nζ)\cos(2N\zeta) radial modulations in Eq. (16). In Family 1 (m1=1m_1=1) with ρa=ρ0\rho_a=\rho_0, C2(ρa)=0C_2(\rho_a)=0 eliminates the cos⁡(Nζ)\cos(N\zeta) term in Eq. (16), resulting in

Raxis(ζ)=ρ01−2cos⁡(2Nζ)(2N+1)2+1(2N+1)4,zaxis(ζ)=2ρ0N2N+1sin⁡(Nζ).R_{\mathrm{axis}}(\zeta)=\rho_0\sqrt{1-\frac{2\cos(2N\zeta)}{(2N+1)^2}+\frac{1}{(2N+1)^4}},\qquad z_{\mathrm{axis}}(\zeta)=\frac{2\rho_0}{N\sqrt{2N+1}}\sin(N\zeta).

As shown in Figure 4a–Figure 4c, Raxis(ζ)R_{\mathrm{axis}}(\zeta) has 2N2N small oscillations around ρ0=1\rho_0=1 while zaxis(ζ)z_{\mathrm{axis}}(\zeta) has NN vertical oscillations. By contrast, in Families 2 and 3 (as well as Family 1 with ρa>ρ0\rho_a>\rho_0), C2(ρa)≠0C_2(\rho_a)\ne0 retains the cos⁡(Nζ)\cos(N\zeta) term in Eq. (16), resulting in much larger radial oscillations (Figure 4d–Figure 4i). Across all three families of equilibria in Figure 4, both the radial and vertical oscillation amplitudes decrease monotonically with increasing NN.

Figure 4. Three-dimensional magnetic axis curves x⃗axis(ζ)\vec{x}_{\mathrm{axis}}(\zeta), radial profiles Raxis(ζ)R_{\mathrm{axis}}(\zeta), and vertical profiles zaxis(ζ)z_{\mathrm{axis}}(\zeta) across NN for (a)–(c) Family 1 (ρa=ρ0\rho_a = \rho_0), (d)–(f) Family 2 (ι=N\iota= N, ρa=1.12,1.06,1.04\rho_a = 1.12, 1.06, 1.04), and (g)–(i) Family 3 (ι=N\iota= N, ρa=0.97,0.89,0.88\rho_a = 0.97, 0.89, 0.88).

Toroidal volume and volume-averaged βV\beta_V

The volume V(δ)V(\delta) enclosed by ψ=δ\psi=\delta and the volume-averaged plasma beta βV(δ)\beta_V(\delta) are

V(δ)=∫0δV′(ψ) dψ,βV(δ):=2⟨p⟩V⟨∣B⃗∣2⟩V=2∫0δ(p(δ)+2(δ−ψ))V′(ψ) dψ∫0δ(H(ρa,0)+2ψ)V′(ψ) dψ,V(\delta)=\int_0^\delta V'(\psi)\,d\psi,\qquad\beta_V(\delta):=\frac{2\langle p\rangle_V}{\langle|\vec{B}|^2\rangle_V}=\frac{2\int_0^\delta\left(p(\delta)+2(\delta-\psi)\right)V'(\psi)\,d\psi}{\int_0^\delta\left(H(\rho_a,0)+2\psi\right)V'(\psi)\,d\psi},
V′(ψ):=2π∫02πJψθζ(ψ,θ) dθ,Jψθζ(ψ,θ):=∣J(ρ,η)∣∣m2∣∣dr/dρ∣∣C4(ρ)∣,V'(\psi):=2\pi\int_0^{2\pi}J_{\psi\theta\zeta}(\psi,\theta)\,d\theta,\qquad J_{\psi\theta\zeta}(\psi,\theta):=\frac{|\mathcal{J}(\rho,\eta)|}{|m_2||d r/d\rho||C_4(\rho)|},

where the volume average of a quantity ff is

⟨f⟩V:=1V(δ)∫0≤ψ≤δf dV.\langle f\rangle_V:=\frac{1}{V(\delta)}\int_{0\leq\psi\leq\delta}f\,dV.

Figure 5 shows the enclosed volume V(δ)V(\delta) and volume-averaged plasma beta βV(δ)\beta_V(\delta) versus the boundary flux δ\delta across all three families of equilibria for N∈{3,4,5}N\in\{3,4,5\}. For Families 2 and 3, we use the same ρa\rho_a values as in Figure 4, which approximately maximize the peak βV\beta_V for each NN. In all cases, both quantities grow nearly linearly with δ\delta, and V(δ)V(\delta) decreases with NN at fixed δ\delta. Because the ι=±N\iota=\pm N branch of Family 2 has thin cross-sections (Figure 3c) and breaks down at a smaller boundary flux δ\delta, it is limited to a small plasma volume and βV\beta_V. By contrast, Family 3 sustains nested flux surfaces up to a larger boundary flux δ\delta, reaching a larger maximum plasma volume and βV\beta_V at ι=±N\iota=\pm N.

Enclosed plasma volume $V(\delta)$

(a)

(b)

Figure 5. Enclosed plasma volume V(δ)V(\delta) and volume-averaged beta βV(δ)\beta_V(\delta) versus boundary flux δ\delta across N∈{3,4,5}N \in\{3,4,5\} for Family 1 with ι=0\iota= 0 and ρa=ρ0\rho_a = \rho_0, Family 2 with ι=±N\iota= \pm N, and Family 3 with ι=±N\iota= \pm N. Markers indicate the maximum boundary flux.

Ideal-MHD interchange stability

Because every field line closes after one toroidal transit with ∮∣B⃗∣−1 dℓ=2π\oint|\vec{B}|^{-1}\,d\ell=2\pi, the enclosed toroidal magnetic flux is Φt(ψ)=V(ψ)/2π\Phi_t(\psi)=V(\psi)/2\pi and d2V/dΦt2=0d^2V/d\Phi_t^2=0, meaning there is neither magnetic shear nor a stabilizing magnetic well (d2V/dΦt2<0d^2V/d\Phi_t^2<0) or magnetic hill (d2V/dΦt2>0d^2V/d\Phi_t^2>0). To determine stability, we therefore evaluate the ideal-MHD energy principle [28, 29] directly,

under which an equilibrium is unstable if δW(ξ⃗)<0\delta W(\vec{\xi}) < 0 for any smooth displacement ξ⃗\vec{\xi} vanishing on the plasma boundary ψ=δ\psi= \delta:

δW(ξ⃗)=12∫0≤ψ≤δ[∣Q⃗∣2+J⃗⋅(ξ⃗×Q⃗)+(ξ⃗⋅∇p)(∇⋅ξ⃗)+γp(∇⋅ξ⃗)2] dV,(17)\delta W(\vec{\xi}) = \frac{1}{2}\int_{0 \leq\psi\leq\delta}\left[|\vec{Q}|^2+\vec{J}\cdot(\vec{\xi}\times\vec{Q})+(\vec{\xi}\cdot\nabla p)(\nabla\cdot\vec{\xi})+\gamma p(\nabla\cdot\vec{\xi})^2\right]\,\mathrm{d}V, \tag*{(17)}

where Q⃗:=∇×(ξ⃗×B⃗)\vec{Q} := \nabla\times(\vec{\xi}\times\vec{B}) is the magnetic field perturbation and γ>0\gamma> 0 is the adiabatic index. Setting χ(ρ,η):=(δ−ψ)3η\chi(\rho,\eta) := (\delta-\psi)^3\eta, we consider a trial displacement ξ⃗=ξ⃗0+ϵξ⃗1\vec{\xi} = \vec{\xi}_0+\epsilon\vec{\xi}_1 [30] that vanishes on ψ=δ\psi= \delta, with constant ϵ\epsilon and

ξ⃗0=1J(∂χ∂η∂x⃗∂ρ−∂χ∂ρ∂x⃗∂η),ξ⃗1=χJ(∂p∂η∂x⃗∂ρ−∂p∂ρ∂x⃗∂η).\vec{\xi}_0=\frac{1}{\mathcal{J}}\left(\frac{\partial\chi}{\partial\eta}\frac{\partial\vec{x}}{\partial\rho}-\frac{\partial\chi}{\partial\rho}\frac{\partial\vec{x}}{\partial\eta}\right),\qquad \vec{\xi}_1=\frac{\chi}{\mathcal{J}}\left(\frac{\partial p}{\partial\eta}\frac{\partial\vec{x}}{\partial\rho}-\frac{\partial p}{\partial\rho}\frac{\partial\vec{x}}{\partial\eta}\right).

Because J(ρ,η)\mathcal{J}(\rho,\eta) is independent of ζ\zeta and B⃗=∂ζx⃗=J(∇η×∇ρ)\vec{B}=\partial_\zeta\vec{x}=\mathcal{J}(\nabla\eta\times\nabla\rho), the pure interchange ξ⃗0\vec{\xi}_0 satisfies ξ⃗0×B⃗=∇χ\vec{\xi}_0\times\vec{B}=\nabla\chi; it therefore neither perturbs the magnetic field nor compresses the plasma, resulting in δW(ξ⃗0)=0\delta W(\vec{\xi}_0)=0. Adding the flux-surface-tangent displacement ϵξ⃗1\epsilon\vec{\xi}_1 introduces parallel magnetic and plasma compression governed by

g(ρ,η):=2(δ−ψ)3J∂ψ∂ρ,g(\rho,\eta):=\frac{2(\delta-\psi)^3}{\mathcal{J}}\frac{\partial\psi}{\partial\rho},

such that Q⃗=−ϵgB⃗\vec{Q}=-\epsilon g\vec{B}, ∇⋅ξ⃗=ϵg\nabla\cdot\vec{\xi}=\epsilon g, and ξ⃗⋅∇p=−g\vec{\xi}\cdot\nabla p=-g. Using J⃗⋅(ξ⃗×Q⃗)=ϵgξ⃗⋅(J⃗×B⃗)=−ϵg2\vec{J}\cdot(\vec{\xi}\times\vec{Q})=\epsilon g\vec{\xi}\cdot(\vec{J}\times\vec{B})=-\epsilon g^2, Eq. (17) simplifies to

δW(ξ⃗0+ϵξ⃗1)=12∫0≤ψ≤δg2[−2ϵ+ϵ2(∣B⃗∣2+γp)] dV.\delta W(\vec{\xi}_0+\epsilon\vec{\xi}_1)=\frac{1}{2}\int_{0\leq\psi\leq\delta}g^2\left[-2\epsilon+\epsilon^2\left(|\vec{B}|^2+\gamma p\right)\right]\,\mathrm{d}V.

which is negative for sufficiently small ϵ>0\epsilon> 0, proving that all three families of equilibria are ideal-MHD interchange unstable.

Exterior vacuum field

To benchmark free-boundary equilibrium codes (e.g., VMEC [31] and DESC [32]), we construct a vacuum magnetic field B⃗vac\vec{B}_{\mathrm{vac}} in a thin layer outside the plasma boundary ψ=δ\psi=\delta that matches the equilibrium magnetic field B⃗\vec{B} across the interface:

∇×B⃗vac=0⃗and∇⋅B⃗vac=0for δ<ψ≤δvac,B⃗vac∣ψ=δ=B⃗∣ψ=δ,(18)\nabla\times\vec{B}_{\mathrm{vac}}=\vec{0}\quad\text{and}\quad\nabla\cdot\vec{B}_{\mathrm{vac}}=0\quad\text{for }\delta<\psi\leq\delta_{\mathrm{vac}},\qquad\left.\vec{B}_{\mathrm{vac}}\right|_{\psi=\delta}=\left.\vec{B}\right|_{\psi=\delta}, \tag*{(18)}

where δvac>δ\delta_{\mathrm{vac}} > \delta is close enough to δ\delta that the coordinate map (3) remains invertible. Matching all components of B⃗\vec{B} implies no surface current, so pressure balance requires p(δ)=0p(\delta) = 0, i.e., p0=2δp_0 = 2\delta.

For Family 1 with ρa=ρ0\rho_a = \rho_0, expressing B⃗=∂ζx⃗\vec{B} = \partial_\zeta\vec{x} in the reciprocal basis (∇ρ,∇ζ,∇η)(\nabla\rho, \nabla\zeta, \nabla\eta) yields

B⃗=H(ρ,η)∇ζ+∇Φ(ρ,ζ,η),(19)\vec{B} = H(\rho,\eta)\nabla\zeta+ \nabla\Phi(\rho,\zeta,\eta), \tag*{(19)}

where Φ(ρ,ζ,η)\Phi(\rho,\zeta,\eta) is the closed-form, single-valued scalar potential

Φ(ρ,ζ,η):=2(N+1)N(C1(ρ)+(2N+1)C3(ρ))ηC4(ρ)cos⁡(Nζ).(20)\Phi(\rho,\zeta,\eta) := \frac{2(N+1)}{N}\left(C_1(\rho) + (2N+1)C_3(\rho)\right)\eta C_4(\rho)\cos(N\zeta). \tag*{(20)}

The exterior vacuum magnetic field is

B⃗vac=(H(ρa,0)+2δ)∇ζ+∇Φ(ρ,ζ,η)+∇[∑n=2∞αn(ρ,ζ,η)(ψ(ρ,η)−δ)n],(21)\vec{B}_{\mathrm{vac}} = \left(H(\rho_a,0) + 2\delta\right)\nabla\zeta+ \nabla\Phi(\rho,\zeta,\eta) + \nabla\left[\sum_{n=2}^{\infty}\alpha_n(\rho,\zeta,\eta)\left(\psi(\rho,\eta)-\delta\right)^n\right], \tag*{(21)}

where the coefficients αn(ρ,ζ,η)\alpha_n(\rho,\zeta,\eta) are determined purely by differentiation via the recurrence

α2=∇ψ⋅∇ζ∣∇ψ∣2,α3=∇2ζ−2∇ψ⋅∇α2−(∇2ψ)α23∣∇ψ∣2,αn=−2∇ψ⋅∇αn−1+(∇2ψ)αn−1+1n−1∇2αn−2n∣∇ψ∣2for n≥4.(22)\begin{aligned} \alpha_2 &= \frac{\nabla\psi\cdot\nabla\zeta}{\lvert\nabla\psi\rvert^2}, \qquad \alpha_3 = \frac{\nabla^2\zeta- 2\nabla\psi\cdot\nabla\alpha_2 - (\nabla^2\psi)\alpha_2}{3\lvert\nabla\psi\rvert^2}, \\ \alpha_n &= -\frac{2\nabla\psi\cdot\nabla\alpha_{n-1} + (\nabla^2\psi)\alpha_{n-1} + \frac{1}{n-1}\nabla^2\alpha_{n-2}}{n\lvert\nabla\psi\rvert^2} \qquad\text{for } n \ge4. \tag*{(22)} \end{aligned}

To verify that Eqs. (21)–(22) satisfy Eq. (18), note first that ∇×B⃗vac=0⃗\nabla\times\vec{B}_{\mathrm{vac}} = \vec{0} holds identically because every term in Eq. (21) is locally a gradient. Second, subtracting Eq. (19) from Eq. (21) gives

B⃗vac−B⃗=−2(ψ(ρ,η)−δ)∇ζ+∇[∑n=2∞αn(ρ,ζ,η)(ψ(ρ,η)−δ)n],(23)\vec{B}_{\mathrm{vac}} - \vec{B} = -2(\psi(\rho,\eta)-\delta)\nabla\zeta+ \nabla\left[\sum_{n=2}^{\infty}\alpha_n(\rho,\zeta,\eta)(\psi(\rho,\eta)-\delta)^n\right], \tag*{(23)}

where all terms vanish on ψ=δ\psi= \delta, confirming the boundary condition in Eq. (18). Third, taking the divergence of Eq. (23) and using ∇⋅B⃗=0\nabla\cdot\vec{B}=0 from Eq. (9) yields

∇⋅B⃗vac=−2∇ψ⋅∇ζ−2(ψ−δ)∇2ζ+∇2[∑n=2∞αn(ψ−δ)n],\nabla\cdot\vec{B}_{\mathrm{vac}} = -2\nabla\psi\cdot\nabla\zeta- 2(\psi-\delta)\nabla^2\zeta+ \nabla^2\left[\sum_{n=2}^{\infty}\alpha_n(\psi-\delta)^n\right],

whose coefficients of each power of ψ−δ\psi-\delta vanish by the recurrence relation in Eq. (22). Truncating the series at n=nmax⁡n=n_{\max} preserves ∇×B⃗vac=0⃗\nabla\times\vec{B}_{\mathrm{vac}}=\vec{0} and B⃗vac=B⃗\vec{B}_{\mathrm{vac}}=\vec{B} on ψ=δ\psi=\delta exactly, while leaving ∇⋅B⃗vac=O((ψ−δ)nmax⁡−1)\nabla\cdot\vec{B}_{\mathrm{vac}}=O((\psi-\delta)^{n_{\max}-1}), as verified in Figure 6a. For N=3N=3, δ=0.02\delta=0.02, and nmax⁡=10n_{\max}=10, Figure 6b and Figure 6c show the divergence residual ∣∇⋅B⃗vac∣/∣B⃗vac∣\lvert\nabla\cdot\vec{B}_{\mathrm{vac}}\rvert/\lvert\vec{B}_{\mathrm{vac}}\rvert and the continuous magnetic field strength across ψ=δ\psi=\delta in the ϕ=0\phi=0 cross-section for δvac=1.2δ\delta_{\mathrm{vac}}=1.2\delta.

Figure 6. Verification of the truncated exterior vacuum field in Eq. (21) for Family 1 with N=3N = 3, ρa=ρ0\rho_a = \rho_0, and plasma boundary δ=0.02\delta= 0.02. (a) Normalized divergence residual along θ=π/4\theta= \pi/4 and ζ=π/4N\zeta= \pi/4N versus distance ψ−δ\psi- \delta from the boundary across truncation orders nmax⁡n_{\max}. (b) Poloidal cross-section at ϕ=0\phi= 0 showing the normalized divergence residual for nmax⁡=10n_{\max} = 10. (c) Poloidal cross-section at ϕ=0\phi= 0 showing the magnetic field magnitude ∣B⃗∣|\vec{B}| in the interior plasma (0≤ψ≤δ0 \leq\psi\leq\delta) and ∣B⃗vac∣|\vec{B}_{\mathrm{vac}}| in the exterior shell (δ≤ψ≤δvac=1.2δ\delta\leq\psi\leq\delta_{\mathrm{vac}} = 1.2\delta) for nmax⁡=10n_{\max} = 10.

Note that for ψ>δ\psi>\delta, ψ(ρ,η)\psi(\rho,\eta) serves only as a spatial coordinate rather than a flux function of B⃗vac\vec{B}_{\mathrm{vac}}, since B⃗vac⋅∇ψ=O((ψ−δ)2)≠0\vec{B}_{\mathrm{vac}}\cdot\nabla\psi=O((\psi-\delta)^2)\ne0 and exterior field lines cross surfaces of constant ψ\psi. Nevertheless, because B⃗vac\vec{B}_{\mathrm{vac}} inherits both discrete NN-fold rotational symmetry (2) and reflection symmetry (11) across ϕ∗=π/2N\phi_*=\pi/2N, every exterior field line near ψ=δ\psi=\delta closes after one toroidal transit (ι=0\iota=0). Consequently, prescribing ψvac=ψ\psi_{\mathrm{vac}}=\psi on a poloidal cross-section and integrating B⃗vac⋅∇ψvac=0\vec{B}_{\mathrm{vac}}\cdot\nabla\psi_{\mathrm{vac}}=0 along the closed field lines defines a smooth, single-valued function ψvac(x⃗)=ψ(ρ,η)+O((ψ−δ)2)\psi_{\mathrm{vac}}(\vec{x})=\psi(\rho,\eta)+O((\psi-\delta)^2) whose level sets form nested toroidal magnetic surfaces. Where the series in Eq. (21) converges, these level sets are exact flux surfaces, whereas truncating at n=nmax⁡n=n_{\max} preserves B⃗vac⋅∇ψvac=0\vec{B}_{\mathrm{vac}}\cdot\nabla\psi_{\mathrm{vac}}=0 while conserving magnetic flux up to O((ψ−δ)nmax⁡−1)O((\psi-\delta)^{n_{\max}-1}).

In free-boundary benchmarks, B⃗\vec{B} and B⃗vac\vec{B}_{\mathrm{vac}} can be used either by prescribing B⃗vac⋅n^\vec{B}_{\mathrm{vac}}\cdot\hat{n} on a control surface δ<ψ<δvac\delta<\psi<\delta_{\mathrm{vac}} (e.g. ψ=1.2δ\psi=1.2\delta in Figure 6b and Figure 6c), or by supplying the external field B⃗ext=B⃗vac−B⃗p\vec{B}_{\mathrm{ext}}=\vec{B}_{\mathrm{vac}}-\vec{B}_p, where B⃗p\vec{B}_p is the Biot–Savart field of the plasma current J⃗=∇×B⃗=2∇ψ×∇ζ\vec{J}=\nabla\times\vec{B}=2\nabla\psi\times\nabla\zeta. For the remaining equilibria with C2(ρ)≠0C_2(\rho)\ne0, the plasma carries a net toroidal current, so a matching curl-free field must also include a multivalued poloidal potential. Constructing this potential requires the radial extent of the boundary cross-section, which in general must be found numerically by root finding, so we restrict the vacuum field to Family 1 with ρa=ρ0\rho_a=\rho_0.

Discussion

This paper introduces three closed-form families of non-axisymmetric ideal MHD equilibria with nested toroidal flux surfaces, stellarator symmetry, and non-constant pressure for any integer number of field periods N≥1N \geq1. Lortz [9] previously proved the existence of ι=0\iota= 0 toroidal equilibria in the limit of low plasma beta via an iterative scheme, relying on reflection symmetry. Family 1 with ρa=ρ0\rho_a = \rho_0 provides an explicit realization of the Lortz [9] equilibria in elementary functions that guarantees nested toroidal flux surfaces, holds at finite plasma beta for arbitrary N≥1N \geq1. Relaxing reflection symmetry yields additional ι=0\iota= 0 solutions in Family 1 and Family 2 (N≥2N \geq2), as well as equilibria with nonzero rotational transform ι=±N\iota= \pm N in Family 2 and Family 3 for arbitrary N≥3N \geq3.

The present equilibria are the first closed-form solutions to provide benchmarks for verifying three-dimensional MHD equilibrium solvers in the N≥3N \geq3 regime. For Family 1 with ρa=ρ0\rho_a = \rho_0, the matched exterior vacuum field of § 5 further extends this benchmark to free-boundary equilibrium codes. Now that several distinct classes of exact non-axisymmetric toroidal equilibria have been found, there are likely many more waiting to be discovered. It would be interesting to investigate whether a more general mathematical theory governs and unifies these classes of exact three-dimensional solutions.

Acknowledgments on AI assistance

This research was conducted with the assistance of Gemini 4 multi-agent workflows within the Google Antigravity environment. The authors posed the problem of discovering toroidal equilibria for N≥3N \geq3 and guided the agents through the analysis. Gemini also assisted in preparing the manuscript and the accompanying SymPy verification and plotting scripts. All mathematical derivations and symbolic checks were reviewed and validated by the authors, who assume full responsibility for the manuscript.

Origin of the equilibria

Exploiting symmetries to restrict the form of a solution ansatz is a standard route to exact solutions of differential equations [33, 34], including closed-form MHD equilibria [35], and symbolic packages are routinely used to automate the resulting calculations [36, 37, 38]. In that spirit, we outline how the equilibria in § 2 were found by perturbing the axisymmetric coordinate map and imposing discrete NN-fold rotational symmetry (2) to derive Eqs. (3)–(8) from the ideal MHD system (1).

Covariant field-line formulation

We formulate the problem as an inverse coordinate map x⃗(ρ,ζ,η)\vec{x}(\rho,\zeta,\eta) with Jacobian determinant

J=det⁡(∂x⃗∂(ρ,ζ,η)).\mathcal{J}=\det\left(\frac{\partial\vec{x}}{\partial(\rho,\zeta,\eta)}\right).

Aligning the magnetic field with the ζ\zeta-coordinate curves via B⃗=∂ζx⃗\vec{B}=\partial_{\zeta}\vec{x}, the directional derivative along B⃗\vec{B} is B⃗⋅∇=∂ζ\vec{B}\cdot\nabla=\partial_{\zeta}, and the divergence-free condition reduces to

∇⋅B⃗=1J∂J∂ζ=0⟺∂J∂ζ=0.(24)\nabla\cdot\vec{B}=\frac{1}{\mathcal{J}}\frac{\partial\mathcal{J}}{\partial\zeta}=0\Longleftrightarrow\frac{\partial\mathcal{J}}{\partial\zeta}=0. \tag*{(24)}

We first rewrite J⃗×B⃗\vec{J}\times\vec{B} using the vector identity

J⃗×B⃗=(B⃗⋅∇)B⃗−12∇∣B⃗∣2=∂2x⃗∂ζ2−12∇(∣∂x⃗∂ζ∣2).\vec{J}\times\vec{B}=(\vec{B}\cdot\nabla)\vec{B}-\frac{1}{2}\nabla|\vec{B}|^2=\frac{\partial^2\vec{x}}{\partial\zeta^2}-\frac{1}{2}\nabla\left(\left|\frac{\partial\vec{x}}{\partial\zeta}\right|^2\right).

To express the force balance equation J⃗×B⃗=∇p\vec{J}\times\vec{B}=\nabla p in the field-aligned coordinates, we take its dot product with each of the tangent vectors ∂x⃗∂ρ\frac{\partial\vec{x}}{\partial\rho}, ∂x⃗∂η\frac{\partial\vec{x}}{\partial\eta}, and ∂x⃗∂ζ\frac{\partial\vec{x}}{\partial\zeta}, which yields

∂x⃗∂ρ⋅(J⃗×B⃗)=∂gρζ∂ζ−∂gζζ∂ρ=∂p∂ρ,(25)\frac{\partial\vec{x}}{\partial\rho}\cdot(\vec{J}\times\vec{B})=\frac{\partial g_{\rho\zeta}}{\partial\zeta}-\frac{\partial g_{\zeta\zeta}}{\partial\rho}=\frac{\partial p}{\partial\rho}, \tag*{(25)}
∂x⃗∂η⋅(J⃗×B⃗)=∂gηζ∂ζ−∂gζζ∂η=∂p∂η,(26)\frac{\partial\vec{x}}{\partial\eta}\cdot(\vec{J}\times\vec{B})=\frac{\partial g_{\eta\zeta}}{\partial\zeta}-\frac{\partial g_{\zeta\zeta}}{\partial\eta}=\frac{\partial p}{\partial\eta}, \tag*{(26)}
∂x⃗∂ζ⋅(J⃗×B⃗)=∂gζζ∂ζ−∂gζζ∂ζ=0=∂p∂ζ,(27)\frac{\partial\vec{x}}{\partial\zeta}\cdot(\vec{J}\times\vec{B})=\frac{\partial g_{\zeta\zeta}}{\partial\zeta}-\frac{\partial g_{\zeta\zeta}}{\partial\zeta}=0=\frac{\partial p}{\partial\zeta}, \tag*{(27)}

where the metric components are

gρζ=∂x⃗∂ρ⋅∂x⃗∂ζ,gηζ=∂x⃗∂η⋅∂x⃗∂ζ,gζζ=∣∂x⃗∂ζ∣2=∣B⃗∣2.(28)g_{\rho\zeta}=\frac{\partial\vec{x}}{\partial\rho}\cdot\frac{\partial\vec{x}}{\partial\zeta},\qquad g_{\eta\zeta}=\frac{\partial\vec{x}}{\partial\eta}\cdot\frac{\partial\vec{x}}{\partial\zeta},\qquad g_{\zeta\zeta}=\left|\frac{\partial\vec{x}}{\partial\zeta}\right|^2=|\vec{B}|^2. \tag*{(28)}

For a ζ\zeta-independent pressure p(ρ,η)p(\rho,\eta), Eq. (27) is satisfied automatically, reducing the three-dimensional MHD system (1) to Eqs. (24)–(26).

Coordinate ansatz

Starting from the axisymmetric coordinate mapping x+iy=ρeiζx+iy=\rho e^{i\zeta}, z=ηz=\eta, we add an NN-periodic vertical oscillation

z(ρ,ζ,η)=η+C5(ρ,η)sin⁡(Nζ),z(\rho,\zeta,\eta)=\eta+C_5(\rho,\eta)\sin(N\zeta),

resulting in the metric (28) components

gρζ=∂x∂ρ∂x∂ζ+∂y∂ρ∂y∂ζ+NC52∂C5∂ρsin⁡(2Nζ),g_{\rho\zeta}=\frac{\partial x}{\partial\rho}\frac{\partial x}{\partial\zeta}+\frac{\partial y}{\partial\rho}\frac{\partial y}{\partial\zeta}+\frac{NC_5}{2}\frac{\partial C_5}{\partial\rho}\sin(2N\zeta),
gηζ=∂x∂η∂x∂ζ+∂y∂η∂y∂ζ+NC5cos⁡(Nζ)+NC52∂C5∂ηsin⁡(2Nζ),g_{\eta\zeta}=\frac{\partial x}{\partial\eta}\frac{\partial x}{\partial\zeta}+\frac{\partial y}{\partial\eta}\frac{\partial y}{\partial\zeta}+NC_5\cos(N\zeta)+\frac{NC_5}{2}\frac{\partial C_5}{\partial\eta}\sin(2N\zeta),
gζζ=(∂x∂ζ)2+(∂y∂ζ)2+N2C522(1+cos⁡(2Nζ)).g_{\zeta\zeta}=\left(\frac{\partial x}{\partial\zeta}\right)^2+\left(\frac{\partial y}{\partial\zeta}\right)^2+\frac{N^2C_5^2}{2}\left(1+\cos(2N\zeta)\right).

Since p(ρ,η)p(\rho,\eta) is independent of ζ\zeta, the resulting harmonics NN and 2N2N in Eqs. (25)–(26) must be cancelled by matching harmonics in x+iyx+iy.

The NN-fold symmetry [2] restricts x+iyx+iy to the Fourier series

x+iy=eiζ∑nFn(ρ,η)einNζ=∑nFn(ρ,η)ei(1+nN)ζ.x+iy=e^{i\zeta}\sum_n F_n(\rho,\eta)e^{inN\zeta}=\sum_n F_n(\rho,\eta)e^{i(1+nN)\zeta}.

The only set of modes producing beat frequencies NN and 2N2N without higher ones is three consecutive harmonics n∈{k−1,k,k+1}n\in\{k-1,k,k+1\} centered at an integer index kk. For the magnetic axis to wind once around the zz-axis, the fundamental mode n=0n=0 must be one of these three harmonics. Choosing n=0n=0 to be the first, second, or third harmonic gives k=1k=1, 00, or −1-1, which defines the three families of equilibria with mode numbers in Eq. (4).

Cancelling the 2Nζ2N\zeta harmonics in Eqs. (24) and (26) allows Fk±1F_{k\pm1} and C5C_5 to be independent of η\eta, so only FkF_k depends on η\eta. Taking FkF_k to be linear in η\eta, we parameterize the amplitudes in terms of real functions C1(ρ),…,C5(ρ)C_1(\rho),\ldots,C_5(\rho) as

Fk−1(ρ)=C1(ρ),Fk(ρ,η)=C2(ρ)+iηC4(ρ),Fk+1(ρ)=−C3(ρ),F_{k-1}(\rho)=C_1(\rho),\qquad F_k(\rho,\eta)=C_2(\rho)+i\eta C_4(\rho),\qquad F_{k+1}(\rho)=-C_3(\rho),

which yields the coordinate ansatz (3):

x+iy=C1(ρ)eim1ζ+(C2(ρ)+iηC4(ρ))eim2ζ−C3(ρ)eim3ζ,z=η+C5(ρ)sin⁡(Nζ).(29)\begin{aligned} x+iy&=C_1(\rho)e^{im_1\zeta}+\left(C_2(\rho)+i\eta C_4(\rho)\right)e^{im_2\zeta}-C_3(\rho)e^{im_3\zeta},\\ z&=\eta+C_5(\rho)\sin(N\zeta). \tag*{(29)} \end{aligned}

Determining the auxiliary functions C1,…,C5C_1,\ldots,C_5

Substituting the ansatz (29) into Eqs. (24)–(26), the coefficients of each ζ\zeta-harmonic must vanish independently, while the ζ\zeta-independent terms determine the pressure p(ρ,η)p(\rho,\eta). We solve these conditions in three steps:

  1. Second harmonics (2Nζ)(2N\zeta). Setting the 2Nζ2N\zeta coefficients in Eqs. (24)–(26) to zero yields the Euler differential equations

ρdC1dρ−m1C1(ρ)=0,ρdC3dρ+m3C3(ρ)=0,ρdC5dρ+NC5(ρ)=0,\rho\frac{\mathrm{d}C_1}{\mathrm{d}\rho}-m_1C_1(\rho)=0,\qquad\rho\frac{\mathrm{d}C_3}{\mathrm{d}\rho}+m_3C_3(\rho)=0,\qquad\rho\frac{\mathrm{d}C_5}{\mathrm{d}\rho}+NC_5(\rho)=0,

after choosing the radial coordinate ρ\rho to fix the first power law. Integrating directly gives

C1(ρ)=c1ρm1,C3(ρ)=c3ρm3,C5(ρ)=c5ρN.(30)C_1(\rho)=c_1\rho^{m_1},\qquad C_3(\rho)=\frac{c_3}{\rho^{m_3}},\qquad C_5(\rho)=\frac{c_5}{\rho^N}. \tag*{(30)}

where c1c_1, c3c_3, c5c_5 are real integration constants.

  1. Fundamental harmonics (Nζ)(N\zeta) in force balance. With Eq. (30), setting the NζN\zeta coefficients in the force-balance equations (25)–(26) to zero yields

D+(ρ)C4(ρ)=N2C5(ρ),ρddρ(D−(ρ)C2(ρ))+ND−(ρ)C2(ρ)=0,D_+(\rho)C_4(\rho)=N^2C_5(\rho),\qquad\rho\frac{\mathrm{d}}{\mathrm{d}\rho}\left(D_-(\rho)C_2(\rho)\right)+ND_-(\rho)C_2(\rho)=0,

with D±(ρ)D_\pm(\rho) defined in Eq. (6). Solving the algebraic relation for C4(ρ)C_4(\rho) and integrating the differential equation from ρa\rho_a to ρ\rho for C2(ρ)C_2(\rho) gives

C4(ρ)=N2C5(ρ)D+(ρ),C2(ρ)=C2(ρa)(ρaρ)ND−(ρa)D−(ρ).(31)C_4(\rho)=\frac{N^2C_5(\rho)}{D_+(\rho)},\qquad C_2(\rho)=C_2(\rho_a)\left(\frac{\rho_a}{\rho}\right)^N\frac{D_-(\rho_a)}{D_-(\rho)}. \tag*{(31)}

Divergence-free closure and ζ\zeta-independent terms

Substituting Eqs. (30)–(31) into the remaining NζN\zeta coefficient of Eq. (24) gives the consistency condition

N2c52=4m1m3c1c3>0⟺N2C5(ρ)2=4m1m3C1(ρ)C3(ρ).(32)N^{2}c_{5}^{2}=4m_{1}m_{3}c_{1}c_{3}>0 \quad\Longleftrightarrow\quad N^{2}C_{5}(\rho)^{2}=4m_{1}m_{3}C_{1}(\rho)C_{3}(\rho). \tag*{(32)}

The constants c1c_{1} and c3c_{3} can be fixed without loss of generality in terms of a single length scale ρ0>0\rho_{0}>0, which approximates the major radius:

c1=sign⁡(m1)ρ01−m1m12,c3=ρ01+m3m32,c5=2ρ01+NN∣m1∣m3,c_{1}=\frac{\operatorname{sign}(m_{1})\rho_{0}^{1-m_{1}}}{m_{1}^{2}},\qquad c_{3}=\frac{\rho_{0}^{1+m_{3}}}{m_{3}^{2}},\qquad c_{5}=\frac{2\rho_{0}^{1+N}}{N\sqrt{\lvert m_{1}\rvert m_{3}}},

where c5c_{5} follows from Eq. (32), taking the positive root without loss of generality. The constants above recover Eq. (5). The ζ\zeta-independent terms in Eqs. (25)–(26) then reduce to

∂p∂ρ=−∂H∂ρ,∂p∂η=−∂H∂η,\frac{\partial p}{\partial\rho}=-\frac{\partial H}{\partial\rho},\qquad\frac{\partial p}{\partial\eta}=-\frac{\partial H}{\partial\eta},

with H(ρ,η)H(\rho,\eta) defined in Eq. (8), which integrate to Eq. (8). Setting the radial pressure derivative to zero at the magnetic axis ρ=ρa\rho=\rho_{a} gives Eq. (6).

References

  1. [1]V. D. Shafranov. Plasma equilibrium in a magnetic field. In M. A. Leontovich, editor, Reviews of Plasma Physics, volume 2, pages 103–151. Consultants Bureau, New York, 1966.
  2. [2]L. S. Solov’ev. The theory of hydromagnetic stability of toroidal plasma configurations. Soviet Physics JETP, 26(2):400–407, 1968.
  3. [3]S. B. Zheng, A. J. Wootton, and E. R. Solano. Analytical tokamak equilibrium for shaped plasmas. Physics of Plasmas, 3(3):1176–1178, 1996. doi: 10.1063/1.871772.
  4. [4]A. J. Cerfon and J. P. Freidberg. “One size fits all” analytic solutions to the Grad–Shafranov equation. Physics of Plasmas, 17(3):032502, 2010. doi: 10.1063/1.3328818.
  5. [5]A. H. Boozer. Plasma equilibrium with rational magnetic surfaces. The Physics of Fluids, 24(11):1999–2003, 1981. doi: 10.1063/1.863297.DOI
  6. [6]P. Helander. Theory of plasma confinement in non-axisymmetric magnetic fields. Reports on Progress in Physics, 77(8):087001, 2014. doi: 10.1088/0034-4885/77/8/087001.
  7. [7]H. Grad. Toroidal containment of a plasma. The Physics of Fluids, 10(1):137–154, 1967. doi: 10.1063/1.1761965.DOI
  8. [8]H. Grad. Theory and applications of the nonexistence of simple toroidal plasma equilibrium. International Journal of Fusion Energy, 3(2):33–46, 1985.
  9. [9]D. Lortz. Über die Existenz toroidaler magnetohydrostatischer Gleichgewichte ohne Rotationstransformation. Zeitschrift für angewandte Mathematik und Physik (ZAMP), 21(2):196–211, 1970. doi: 10.1007/BF01590644.
  10. [10]J. Gómez-Serrano, L. Liehr, and M. A. Taylor. Counterexamples to Grad’s conjecture. arXiv:2609.24739, 2026.
  11. [11]M. Landreman. Analytic toroidal 3D MHD equilibria and steady Euler flows with invariant surfaces. arXiv:2609.26742, 2026.
  12. [12]O. Issan, J. Citrin, M. Landreman, and B. Tracey. Analytic non-axisymmetric 3D MHD equilibria without stellarator symmetry. arXiv:2610.07304, 2026.arxiv.org/abs/2610.07304
  13. [13]W. Sengupta, A. Bhattacharjee, T. J. Klotz, A. O. Brown, J. W. Burby, et al. Exact nonsymmetric vacuum magnetic fields with nested toroidal surfaces. arXiv:2610.28466, 2026.arxiv.org/abs/2610.28466
  14. [14]O. E. López. Exact nonsymmetric vacuum magnetic field with nested toroidal flux surfaces and nonzero rotational transform. arXiv:2610.29921, 2026.arxiv.org/abs/2610.29921
  15. [15]R. L. Dewar and S. R. Hudson. Stellarator symmetry. Physica D: Nonlinear Phenomena, 112(1–2):275–280, 1998. doi: 10.1016/S0167-2789(97)00216-9.DOI
  16. [16]M. C. Zarnstorff, L. A. Berry, A. Brooks, E. Fredrickson, G.-Y. Fu, et al. Physics of the compact advanced stellarator NCSX. Plasma Physics and Controlled Fusion, 43(12A):A237–A249, 2001. doi: 10.1088/0741-3335/43/12A/318.DOI
  17. [17]F. S. B. Anderson, A. F. Almagri, D. T. Anderson, P. G. Matthews, J. N. Talmadge, et al. The Helically Symmetric Experiment (HSX) goals, design and status. Fusion Technology, 27(3T):273–277, 1995. doi: 10.13182/FST95-A11947086.DOI
  18. [18]C. Alejaldre, J. Alonso, L. Almoguera, E. Ascasíbar, A. Baciero, et al. First plasmas in the TJ-II flexible heliac. Plasma Physics and Controlled Fusion, 41(3A):A539–A548, 1999. doi: 10.1088/0741-3335/41/3A/047.DOI
  19. [19]G. Grieger, W. Lotz, P. Merkel, J. Nührenberg, J. Sapper, et al. Physics optimization of stellarators. Physics of Fluids B: Plasma Physics, 4(7):2081–2091, 1992. doi: 10.1063/1.860481.DOI
  20. [20]T. Klinger, T. Andreeva, S. Bozhenkov, C. Brandt, R. Burhenn, et al. Overview of first Wendelstein 7-X high-performance operation. Nuclear Fusion, 59(11):112004, 2019. doi: 10.1088/1741-4326/ab03a7.DOI
  21. [21]A. Iiyoshi, A. Komori, A. Ejiri, M. Emoto, H. Funaba, et al. Overview of the Large Helical Device project. Nuclear Fusion, 39(9Y):1245–1256, 1999. doi: 10.1088/0029-5515/39/9Y/313.DOI
  22. [22]S. P. Hirshman and J. C. Whitson. Steepest-descent moment method for three-dimensional magnetohydrodynamic equilibria. The Physics of Fluids, 26(12):3553–3568, 1983. doi: 10.1063/1.864116.DOI
  23. [23]D. W. Dudt and E. Kelemen. DESC: A stellarator equilibrium solver. Physics of Plasmas, 27(10):102513, 2020. doi: 10.1063/5.0020743.DOI
  24. [24]F. Hindelang, G. G. Plunk, and O. Maj. Computing MHD equilibria of stellarators with a flexible coordinate frame. Plasma Physics and Controlled Fusion, 67(4):045002, 2025. doi: 10.1088/1361-6587/adb5f9.
  25. [25]S. R. Hudson, R. L. Dewar, G. Dennis, M. J. Hole, M. McGann, et al. Computation of multi-region relaxed magnetohydrodynamic equilibria. Physics of Plasmas, 19(11):112502, 2012. doi: 10.1063/1.4765691.DOI
  26. [26]W. A. Newcomb. Magnetic differential equations. The Physics of Fluids, 2(4):362–365, 1959. doi: 10.1063/1.1724405.DOI
  27. [27]S. Hamada. Hydromagnetic equilibria and their proper coordinates. Nuclear Fusion, 2(1–2):23–37, 1962. doi: 10.1088/0029-5515/2/1-2/005.DOI
  28. [28]I. B. Bernstein, E. A. Frieman, M. D. Kruskal, and R. M. Kulsrud. An energy principle for hydromagnetic stability problems. Proceedings of the Royal Society of London. Series A, 244(1236):17–40, 1958. doi: 10.1098/rspa.1958.0023.DOI
  29. [29]J. P. Freidberg. Ideal MHD. Cambridge University Press, Cambridge, 2014. doi: 10.1017/CBO9780511795046.DOI
  30. [30]G. O. Spies. Magnetohydrodynamic stability theory with closed magnetic field lines. The Physics of Fluids, 17(2):400–408, 1974. doi: 10.1063/1.1694729.DOI
  31. [31]S. P. Hirshman, W. I. van Rij, and P. Merkel. Three-dimensional free boundary calculations using a spectral Green’s function method. Computer Physics Communications, 43(1):143–155, 1986. doi: 10.1016/0010-4655(86)90058-5.DOI
  32. [32]Rory Conlin, Jonathan Schilling, Daniel W. Dudt, Dario Panici, Rogerio Jorge, and Egemen Kolemen. High order free boundary MHD equilibria in DESC, 2024. URL https://arxiv.org/abs/2412.05680.
  33. [33]P. J. Olver. Applications of Lie Groups to Differential Equations, volume 107 of Graduate Texts in Mathematics. Springer, New York, second edition, 1993. doi: 10.1007/978-1-4612-4350-2.DOI
  34. [34]P. E. Hydon. Symmetry Methods for Differential Equations: A Beginner’s Guide. Number 22 in Cambridge Texts in Applied Mathematics. Cambridge University Press, Cambridge, 2000. doi: 10.1017/CBO9780511623967.DOI
  35. [35]J. M. Keller and A. F. Cheviakov. New exact plasma equilibria with axial and helical symmetry. Physics of Plasmas, 30(9):092308, 2023. doi: 10.1063/5.0164552.
  36. [36]S. Dimas and D. Tsoubelis. SYM: a new symmetry-finding package for Mathematica. In Proceedings of the 10th International Conference in Modern Group Analysis (MOGRAN X), pages 64–70, Nicosia, 2005. University of Cyprus.
  37. [37]A. Meurer, C. P. Smith, M. Paprocki, O. Čertík, S. B. Kirpichev, et al. SymPy: symbolic computing in Python. PeerJ Computer Science, 3:e103, 2017. doi: 10.7717/peerj-cs.103.DOI
  38. [38]F. Oliveri. ReLie: a Reduce program for Lie group analysis of differential equations. Symmetry, 13(10):1826, 2021. doi: 10.3390/sym13101826.DOI

Paper details

vlnc:W_cdhvd3rn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents