In this paper
AbstractExplicit fieldsParaboloidal geometryEllipsoidal geometryVacuum equationsFluxesAxis transformShear at fixed axis transformDomains and checksReferencesWorksPhysics and Astronomy
Exact Nonsymmetric Vacuum Magnetic Fields with Nested Toroidal Surfaces
Abstract
We give a constructive method for exact vacuum magnetic fields whose lines fill nested toroidal surfaces in ordinary space, without continuous spatial symmetry. Complex potential theory on an algebraic curve, combined with doubly periodic traveling waves from soliton theory, gives closed forms for the field, its potential and a surface label. Free parameters set the rotational transform and the magnetic shear, including exactly constant irrational transform. The same fields are exact potential flows with toroidal stream surfaces, and give stringent tests for standard stellarator codes.
Vector fields whose lines fill nested tori appear across physics. They confine plasma in stellarators and tokamaks [1], they are the stream surfaces of steady inviscid flows [2], and they organize the knotted field lines of exact electromagnetic waves [3].
For a vacuum magnetic field with , nesting is not automatic. Without a continuous symmetry, three-dimensional (3D) fields generically develop magnetic islands and chaotic field lines [4]. At the low pressure and small plasma current of present experiments and of reactor start-up, the confining field is essentially the vacuum field of the coils, and the quality of its flux surfaces sets the confinement volume, the edge islands and the starting point for finite-pressure equilibria. Much of the theory of stellarator vacuum fields has therefore been perturbative. Coil currents or shapes are adjusted to cancel the resonant Fourier components that open low-order islands [5]. Quadratic-flux-minimizing and almost-invariant surfaces give integrable fields close to nonintegrable ones [6, 7], and islands in free-boundary equilibria have been removed by targeting resonant harmonics [8]. Pressure-driven currents can shrink or heal vacuum islands at finite [9, 10]. These methods suppress selected islands or construct nearby integrable fields. None supplies a nonsymmetric vacuum field with guaranteed nesting through every rational surface, and so none provides an exact vacuum reference against which the approximations, and the codes built on them, can be tested.
Grad conjectured that smooth equilibria with nested surfaces and pressure require such a symmetry [11]. Counterexamples with pressure appeared only recently [12, 13, 14]. These specific constructions have nonconstant pressure and do not provide vacuum or force-free examples. Earlier exact nonsymmetric vacuum examples live on a flat toroidal quotient rather than in ordinary space [15].
In this Letter we give a constructive method, not an isolated example, that produces exact vacuum fields in 3D configuration space whose lines fill a finite volume with nested toroidal surfaces around an axis with no continuous rigid symmetry (Fig. 1). It combines classical complex potential theory with doubly periodic traveling waves from soliton theory, and it generates infinite families. Its free parameters set the rotational transform , the number of poloidal turns per toroidal turn, and the magnetic shear independently, and they include members with exactly constant . Each field comes in closed form with its potential and an exact surface label. Nesting is therefore proved, not inferred from field-line tracing or from an expansion about the axis. A companion article gives the complete construction, the proofs and the numerical tests [16].
Construction. In two dimensions every harmonic function is the real or imaginary part of an analytic function. Whittaker found the 3D analog: harmonic functions are integrals of analytic functions over a family of complex lines [17]. Bergman restricted the integrand to an algebraic curve in that space [18]. We choose curves for which the construction becomes explicit. The curve fixes a web of confocal quadric surfaces, ellipsoidal or paraboloidal, which serve as curvilinear coordinates through a position map . The same webs carry the classical coordinates in which Laplace’s equation separates into products [19]. Within our paired construction, flatness fixes the web through elliptic functions or their degenerations [16]. Abel’s theorem instead gives two of the resulting harmonic functions a common angle , a toroidal angle that generally differs from the cylindrical azimuth , so that the potential becomes the imaginary part of a sum of an analytic function of and one of . We call the derivatives of these two functions the profiles. The vacuum equations, a 3D generalization of the Cauchy–Riemann equations, then follow from the Cauchy–Riemann equations of the profiles and three identities of the web (Appendix A).
The second ingredient comes from soliton theory. Closing the field on a torus needs profiles periodic in , and a regular magnetic axis needs turning points of the profiles in and , around which the level curves of the surface label close. If both profiles are nonconstant and real on a common section of constant , and the label is conserved whatever the metric of the web, then each profile obeys with a polynomial of degree at most four [16]. These are precisely the degrees for which such an equation has the Painlevé property, with poles as its only movable singularities [20]. The profiles’ Painlevé property and soliton equations thus follow from the existence of the label rather than being assumed. For our even quartic, the solutions are Jacobi elliptic functions, with parameter and [21]. After rescaling, they are periodic traveling waves of the defocusing mKdV equation [22]. These doubly periodic waves meet both requirements. Their imaginary period fits the toroidal circuit, and the second profile is shifted by half an imaginary period, so that the two profiles have distinct turning points that give the axis. Their addition formula, a second use of Abel’s theorem, makes every field line conserve an explicit label . With and the complete elliptic integrals, , , , and , the potential is
The parameters are the axis location , the scale , the phase and , together with the shape parameter of the confocal ellipsoids (Appendix A). Writing for the period of , the period condition with integer makes the field single valued, while gains the constant per toroidal circuit. This is the circulation of a curl-free field around the hole of the torus. With , and , the surface label is
The explicit maps and the Cartesian field for both geometries are given in Appendix A. Landreman’s counterexamples to Grad’s conjecture also carry soliton structure. Along each field line, parametrized by with , a constant multiple of a signed field component is a Jacobi elliptic traveling wave of the focusing mKdV equation [13, 16].
Nested surfaces. In each regular domain the field is analytic, nonvanishing and satisfies
The last inequality makes every line advance around the torus. The label has its strict minimum on the closed curve , which is the magnetic axis, and increases strictly along every outward coordinate ray. Where the position map is one to one, each level is therefore a smooth torus, the levels are nested, and they shrink smoothly onto the axis [16]. This proves nesting through the whole domain, including every rational surface. Rational surfaces carry closed field lines with no islands. The axes are noncircular and isolated, which excludes any continuous rigid symmetry.
Control of transform and shear. Let and be the toroidal and poloidal magnetic fluxes inside the surface . With primes denoting ,
and both fluxes are explicit integrals of the closed-form field, with no fitting of field-line returns. Near the axis the quadratic part of the flux action is the Ermakov–Lewis invariant [23, 24] of the linearized flow in a periodic canonical gauge, and the exact action continues it to finite radius [16]. In the paraboloidal family the two flux derivatives are proportional on every surface, so is exactly constant and irrational for suitable parameters, for example or . The ellipsoidal family has nonzero shear, and can rise or fall outward (Figure 2, left). The axis transform follows from closed-form residue sums (Appendix § 0.1) and is moved continuously by the axis location (Figure 2, right). Exact members with include one whose transform falls by 27% across its domain and one whose transform rises by 2%. A Gardner deformation of the profiles, which turns the mKdV waves into traveling wave solutions of the Gardner equation, changes the axis shear while preserving . Other parameters vary the mirror structure. Nested tori around the axis persist over continuous parameter ranges, and only the extent of the nested region is found member by member (Appendix B).
Figure 2. Independent control of transform and shear. Left: transform magnitude divided by its axis value against normalized toroidal flux for four members. The paraboloidal member has exactly constant irrational transform. The two ellipsoidal members with and profile repetitions have decreasing and increasing transform magnitude, respectively, and the algebraic member has . Right: axis transform against the axis coordinate for the two families. Solid curves are one parameter family each, and dashed curves a second. Colored points mark exact targets , , and , stars mark the members of Figure 1, Figure 3 and Figure 4, labeled by for the ellipsoids, and diamonds the reference members.
Stringent tests for equilibrium codes. The exact outer surface, the toroidal flux, zero pressure and zero toroidal current are sufficient inputs for an equilibrium code. The interior field, the axis and the transform are then predictions that the exact solution checks directly. Figure 3 shows four codes [25, 26, 27, 28] on the paraboloidal member with and on the sheared ellipsoidal member of Figure 1(c,Figure 1d). Away from the axis the VMEC transform error falls with radial resolution to a few times , but next to the axis it stays between and at every tested resolution, the known inaccuracy of VMEC there [25]. VMEC2000 did not converge on the ellipsoidal member. SPEC, given the exact axis as its coordinate axis, reaches errors of order on the paraboloidal and on the ellipsoidal member. DESC, which starts from the exact surfaces, reaches on the paraboloidal and on the ellipsoidal member. The exact solutions measure the accuracy of the computed field directly. These comparisons illustrate how demanding the fields are as tests.
Figure 1. Exact nested magnetic surfaces for the ellipsoidal (top) and paraboloidal (bottom) families, at low rotational transform (left pair) and higher transform (right pair). In (a), (c), (e) and (g) the surfaces enclosing the fractions , and of the toroidal flux are cut away and partly transparent to show the nesting. The thick black curve is the magnetic axis. The navy curve is one field line on the outer surface, drawn slightly outside it for visibility, with an arrow along . It runs nearly parallel to the axis at low and winds visibly at higher . A faint blue plane marks the section shown to the right. Panels (b), (d), (f) and (h) show the same three surfaces (colored) and the surfaces , and (gray) in that plane, at cylindrical azimuth for the ellipsoidal and for the paraboloidal members. Here is the cylindrical radius and + marks the axis.
Figure 3. Rotational transform error of four codes against normalized toroidal flux , for (a) the paraboloidal member with constant and (b) the ellipsoidal member of Figure 1(c,d) with . Upper strips show the exact transform (black) and the finest run of each code. VMEC2000 did not converge in (b) and is omitted. The rise toward is the known VMEC inaccuracy near the magnetic axis [25]. Legends give for VMEC, for SPEC and for DESC.
Particle orbits. Nested magnetic surfaces confine field lines, not particles, which drift across the surfaces. For trapped particles at fixed energy and magnetic moment, bounce averaged drifts follow contours of the action
where is the mirror field and is arclength between the mirror points. Because the field and its derivatives are explicit, can be evaluated to quadrature precision. can be evaluated to quadrature precision. In the ellipsoidal member with and , one trapped class has a closed action contour that crosses magnetic surfaces (Figure 4), an iso-action class in the sense of Ref. [29], which requires only that the drift surface close. Around the launch surface the action varies by 9.4%, yet along the full guiding-center trajectory, over three drift circuits, it varies by only 0.003%. This is a single finite-time orbit. It shows how the exact fields isolate questions of particle confinement from those of field-line confinement, but it does not establish omnigenity.
Figure 4. A trapped guiding-center orbit in the ellipsoidal member with and , at mirror-field depth fraction 0.35 and . (a) Action contours (gray) against the field-line label and the label radius , the square root of the normalized surface label, so that magnetic surfaces are horizontal lines. The closed contour through the launch (orange), 96 turning points of the orbit mapped back along field lines (blue) and the launch magnetic surface (dashed) are shown. (b) Action relative to , its value at the first turning point of the orbit, around the launch magnetic surface (orange) and along the orbit (blue). (c) Action along the orbit against time in axis transit times. The launch is not a turning point. The guiding center drifts about one Larmor radius before its first bounce, so the action of the launch field line is 0.08% below .
Broader implications. The same fields are exact solu-
tions of steady inviscid fluid flow. With velocity and pressure at unit density, they satisfy the incompressible Euler equations with zero vorticity [2]. They are therefore three-dimensional potential flows, the basic model of inviscid aerodynamics, with circulation through the hole of the torus, nested toroidal stream surfaces and no symmetry. In twistor terms, Abel’s theorem on a compact algebraic curve in the space of oriented lines gives the common angle, and the same theorem on the elliptic curve of the soliton profiles gives the surface label [16, 30, 31]. A KdV approach to quasisymmetric magnetic fields selects related KdV and Gardner reductions by requiring the Painlevé property. Deriving that property from quasisymmetry alone remains open [32]. A related complex construction underlies exact electromagnetic fields whose lines form torus knots on nested tori [3]. Our fields are magnetostatic, curl-free analogs, with closed torus-knot field lines on every rational surface and dense lines on every irrational one. Finally, for Hamiltonian dynamics they are explicit integrable systems without symmetry.
In summary, combining complex potential theory with doubly periodic traveling waves from soliton theory yields families of exact vacuum magnetic fields with nested toroidal surfaces in ordinary space, no continuous rigid symmetry, closed-form potentials and labels, and independently controlled transform and shear. Finding external currents that produce a chosen member, and the robustness of the surfaces to perturbations, are design questions for future work.
Under the authors’ direction, OpenAI Codex (recorded model identifiers gpt-6-astra, gpt-6-sol and gpt-6.1-sol) and Anthropic Claude Code (Claude Opus 5.5 and Claude Fable 5.1) assisted with analytical derivations, symbolic and numerical calculations, equilibrium code comparisons, orbit computations, assistance in figure preparation, literature searches and manuscript drafting. The authors set the research questions, selected the examples and reviewed intermediate results and drafts. Checks included analytical identities, symbolic calculations, high-precision evaluations, convergence tests and comparisons between separate agent sessions, as described in the paper. The authors retain responsibility for the scientific claims and the final manuscript. This research was supported by the Simons Foundation/SFARI (Grant No. 560651, AB) and the Department of Energy (Award No. DE-SC0024548).
Explicit fields
With , and , the derivatives of the potential are , and . Both geometries share the Euclidean metric
with and . Writing , the Cartesian field at is
so that .
Paraboloidal geometry
Here and
with and . The member with has , , and , and every surface has .
Ellipsoidal geometry
With the shape parameter , put , and , so that . Then
where functions of and have parameter and those of have parameter . With , and , one has and . The ellipsoidal reference member has , , and , where is the incomplete elliptic integral. At fixed , the cylindrical azimuth of changes monotonically, by per circuit of in the paraboloidal map, where on , and by in the ellipsoidal map.
Vacuum equations
The metric (A1) gives , and . Since and are analytic on the regular domain, their Cauchy–Riemann equations give
The components of along , and are , and , with , so this gives . With the volume element , Eqs. (6) and (9) give
which vanishes because , and in both geometries. For the paraboloidal one, and . In the ellipsoidal one , and depend only on , and , so and .
Fluxes
Write with . On the section the label reduces to and , so the toroidal flux inside the surface is , where is the connected component of that contains the axis point . The poloidal flux crosses the ribbon , , where meets the level , so .
Axis transform
Let and , with nome ratios and such that is an odd integer. Then . For a planar axis, and , the axis transform is with and the finite sum
with . Its single pole entry, and , is replaced by its regular value [16]. Any therefore fixes a closed-form member. The member with in Figure 1 and Figure 3 has and , so , , and . Then and give , and with gives exactly.
Shear at fixed axis transform
For the reference ellipsoid, the Gardner deformation replaces the profiles and by and . The deformed field remains harmonic and has an exact label. On a common finite tube for , its axis transform is unchanged and its axis shear is multiplied by [16].
Domains and checks
The restrictions exclude the poles of the profiles. The maps are one to one and regular for , in the ellipsoidal geometry and for in the paraboloidal geometry. A level is a torus of the family when it stays inside both sets of restrictions over a full circuit, which is certified for each member [16]. The agents identified in the Acknowledgments assisted in developing and running the calculation scripts. The analytical checks, numerical tolerances and comparisons used to assess the results are given here and in the main text. Independent implementations of Eqs. (1), (2) and (5) to (8) at 30 to 40 digits reproduce , on the axis and the stated circulation of to residuals below .
References
- [1]P. Helander, Theory of plasma confinement in non-axisymmetric magnetic fields, Rep. Prog. Phys. 77, 087001 (2014).DOI
- [2]V. I. Arnold and B. A. Khesin, Topological Methods in Hydrodynamics, 2nd ed. (Springer, Cham, 2021).
- [3]H. Kedia, I. Bialynicki-Birula, D. Peralta-Salas, and W. T. M. Irvine, Tying knots in light fields, Phys. Rev. Lett. 111, 150404 (2013).DOI
- [4]J. R. Cary and R. G. Littlejohn, Noncanonical Hamiltonian mechanics and its application to magnetic field line flow, Ann. Phys. 151, 1 (1983).DOI
- [5]J. D. Hanson and J. R. Cary, Elimination of stochasticity in stellarators, Phys. Fluids 27, 767 (1984).
- [6]R. L. Dewar, S. R. Hudson and P. F. Price, Almost invariant manifolds for divergence-free fields, Phys. Rev. Lett. A 194, 49 (1994).
- [7]S. R. Hudson and R. L. Dewar, Construction of an integrable field close to any non-integrable toroidal magnetic field, Phys. Lett. A 247, 246 (1998).DOI
- [8]S. R. Hudson, D. A. Monticello, A. H. Reiman, A. H. Boozer, D. J. Strickler, S. P. Hirshman, and M. C. Zarnstorff, Eliminating islands in high-pressure free-boundary stellarator magnetohydrodynamic equilibrium solutions, Phys. Rev. Lett. 89, 275003 (2002).
- [9]T. Hayashi, T. Sato, P. Merkel, J. Nührenberg and U. Schwenn, Formation and “self-healing” of magnetic islands in finite-β Helias equilibria, Phys. Plasmas 1, 3262 (1994).DOI
- [10]A. Bhattacharjee, T. Hayashi, C. C. Hegna, N. Nakajima and T. Sato, Theory of pressure-induced islands and self-healing in three-dimensional toroidal magnetohydrodynamic equilibria, Phys. Plasmas 2, 883 (1995).
- [11]H. Grad, Toroidal containment of a plasma, Phys. Fluids 10, 137 (1967).DOI
- [12]J. Gómez-Serrano, L. Liehr, and M. A. Taylor, Counterexamples to Grad’s conjecture, arXiv:2609.24739 (2026).
- [13]M. Landreman, Analytic toroidal 3D MHD equilibria and steady Euler flows with invariant surfaces, arXiv:2609.26742v2 (2026).arxiv.org/abs/2609.26742
- [14]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
- [15]H. Weitzner and W. Sengupta, Exact non-symmetric closed line vacuum magnetic fields in a topological torus, Phys. Plasmas 27, 022509 (2020).DOI
- [16]W. Sengupta, A. Bhattacharjee, T. J. Klotz, A. O. Brown, J. W. Burby, V. N. Duarte, S. Patil, and S. Buller, Exact nonsymmetric vacuum magnetic fields with nested toroidal surfaces: Construction, proofs, and tests, in preparation (2026).arxiv.org/abs/2610.28466
- [17]E. T. Whittaker, On the partial differential equations of mathematical physics, Math. Ann. 57, 333 (1903).DOI
- [18]S. Bergmann, Zur Theorie der algebraischen Potentialfunktionen des dreidimensionalen Raumes. II, Math. Ann. 101, 534 (1929).DOI
- [19]L. P. Eisenhart, Separable systems of Stäckel, Ann. Math. 35, 284 (1934).DOI
- [20]A. N. W. Hone, Painlevé tests, singularity structure and integrability, in Integrability, edited by A. V. Mikhailov, Lecture Notes in Physics Vol. 767 (Springer, Berlin, 2009), p. 245.arxiv.org/abs/nlin/0502017
- [21]NIST Digital Library of Mathematical Functions, Chap. 22, https://dlmf.nist.gov/22.
- [22]R. M. Miura, Korteweg–de Vries equation and generalizations. I. A remarkable explicit nonlinear transformation, J. Math. Phys. 9, 1202 (1968).
- [23]H. R. Lewis, Classical and quantum systems with time-dependent harmonic-oscillator-type Hamiltonians, Phys. Rev. Lett. 18, 510 (1967).
- [24]M. P. Perinadin and J. A. Tataronis, Hamiltonian approach to the existence of magnetic surfaces, J. Math. Phys. 26, 2370 (1985).
- [25]D. Panici, R. Conlin, D. W. Dudt, K. Unalmis, and E. Kolemen, The DESC stellarator code suite. Part 1. Quick and accurate equilibria computations, J. Plasma Phys. 89, 955890303 (2023).DOI
- [26]S. P. Hirshman and J. C. Whitson, Steepest-descent moment method for three-dimensional magnetohydrodynamic equilibria, Phys. Fluids 26, 3553 (1983).
- [27]J. Schilling, E. Guirauud, and V. Siska, VMEC++, version 0.8.1 (Proxima Fusion, 2026), doi:10.5281/zenodo.14800157.
- [28]S. R. Hudson, R. L. Dewar, G. Dennis, M. J. Hole, M. McGann, G. von Nessi, and S. Lazerson, Computation of multi-region relaxed magnetohydrodynamic equilibria, Phys. Plasmas 19, 112502 (2012).
- [29]W. Sengupta, A. Bhattacharjee, and S. Buller, Symmetry-agnostic stellarators for collisionless confinement, arXiv:2608.20042 (2026).arxiv.org/abs/2608.20042
- [30]R. Penrose, Twistor algebra, J. Math. Phys. 8, 345 (1967).DOI
- [31]M. Dunajski, Harmonic functions, central quadrics, and twistor theory, Class. Quantum Grav. 20, 3427 (2003).DOI
- [32]W. Sengupta, N. Nikulsin, S. Buller, R. Madan, E. J. Paul, R. Nies, A. A. Kaptanoglu, S. R. Hudson, and A. Bhattacharjee, Periodic Korteweg–de Vries potentials generate quasisymmetric magnetic fields, Phys. Plasmas 33, 092510 (2026).DOI



