# Exact Beltrami flows in a spherical shell

Oleg Bogoyavlenskij

## Abstract

Exact flows of an incompressible fluid satisfying the Beltrami equation inside a spherical shell are constructed in the Cartesian coordinates in terms of elementary functions. Two scale-invariant equations defining two infinite series of eigenvalues λ n and λ ̃ m of the operator curl in the shell with the nonpenetration boundary conditions on the boundary spheres are derived. The corresponding eigenfields are presented in explicit form and their symmetries are investigated. Asymptotics of the eigenvalues λ n and λ ̃ m at n, m → ∞ are obtained.

## 1 Introduction

As known [1], Beltrami equation has many applications to problems of plasma physics and fluid dynamics. Equations of viscous incompressible magnetohydrodynamics have the form [2]

(1.1) V t + curl V × V = grad p ρ + 1 2 | V | 2 + Φ + 1 ρ curl B × B + ν ( t ) Δ V ,

B t = curl ( V × B ) + η ( t ) Δ B , d i v V = 0 , d i v B = 0 .

Here, V(x, t) is fluid velocity, B(x, t) is magnetic field, p(x, t) is the pressure, Φ(x) is the gravitational potential, ρ is constant fluid density, ν(t) is viscosity coefficient, η(t) is magnetic viscosity, and Δ is the Laplace operator.

Equation (1.1) for V(x, t) = 0 reduce to

(1.2) curl B × B = grad p + ρ Φ , B t = 0 , d i v B = 0 .

Equation (1.2) describes plasma equilibria and is equivalent to the generalized Beltrami equation [3]:

(1.3) curl curl B × B = 0 .

The force-free plasma equilibria are defined by condition p(x, t) + Φ(x) = const and are equivalent to the Beltrami equation curl B = λ B, where λ = const in view of equation div B = 0.

If magnetic field B = 0 and viscosity ν(t) = 0, Eq. (1.1) becomes Euler equations for ideal incompressible fluid:

V t + curl V × V = grad p ρ + 1 2 | V | 2 + Φ , d i v V = 0 .

For the steady case and with the additional condition p/ρ + |V|2/2 + Φ = const, the Euler equations reduce to the Beltrami equation

(1.4) curl V ( x ) = λ V ( x ) .

Equation (1.4) implies λ = const in view of div V(x) = 0. Beltrami Eq. (1.4) evidently is invariant with respect to the scale transformations

(1.5) x 1 c x 1 , x 2 c x 2 , x 3 c x 3 , λ 1 c λ ,

where c ≠ 0 is an arbitrary parameter.

In case of collinear vector fields V(x, t) and B(x, t), Eq. (1.1) with nonzero viscosities ν(t) ≠ 0, η(t) ≠ 0 have the following solutions [4]:

(1.6) V ( x , t ) = exp λ 2 t 0 t ν ( τ ) d τ V 1 ( x ) ,

(1.7) B ( x , t ) = C 1 exp λ 2 t 0 t η ( τ ) d τ V 1 ( x ) ,

where V 1(x) is any solution to the Beltrami Eq. (1.4) and pressure p is defined from equation p/ρ + |V|2/2 + Φ = const.

If B(x, t) = 0 then Eq. (1.1) reduce to the Navier–Stokes equations

(1.8) V t + curl V × V = grad p ρ + 1 2 | V | 2 + Φ + ν ( t ) Δ V , d i v V = 0 ,

that have exact solutions (1.6) [4].

In this paper, we construct two infinite series of exact Beltrami flows satisfying Eq. (1.4) inside the spherical shells

(1.9) S R 2 R 1 : R 2 R R 1 , R = x 1 2 + x 2 2 + x 3 2 .

We assume that on the two boundary spheres S 1 2 : R = R 1 and S 2 2 : R = R 2 the nonpenetration boundary condition n(x) ⋅ V(x) = 0 is satisfied. Here, n(x) is vector field of unit normals to the boundary. Since for the spheres S k 2 we have n(x) = x/|x| the boundary conditions have the equivalent form

(1.10) x V n ( x ) = 0

for x S k 2 , k = 1, 2. The same solutions with magnetic field B(x) instead of velocity V(x) describe the force-free magnetic fields inside a spherical shell (1.9). All constructed exact solutions define by formulas (1.6) and (1.7) exact flows of viscous magnetic fluid inside the spherical shell S R 2 R 1 (1.9) with the nonpenetration conditions (1.10).

The main problem consists of finding the eigenvalues λ n of the operator curl and its eigenfields V n (x) that satisfy Eq. (1.4) inside the spherical shell and the nonpenetration condition (1.10) at its two boundaries. Cantarella, DeTurck, Gluck, and Teytel studied in paper [5] vector fields V that “satisfy the equation ∇ × V = λ V , where λ is the eigenvalue of curl having smallest nonzero absolute value among such fields. It is shown that on the ball the energy minimizers are axially symmetric spheromak fields found by Woltjer and Chandrasekhar-Kendall, and on spherical shells they are spheromak-like fields” [5], p. 2766. The authors used the spherical coordinate r, θ, φ in R 3 and the theory of Bessel functions, and the Legendre and Gegenbauer functions applied earlier by Woltjer [6] and by Chandrasekhar and Kendall [7, 8]. The methods of papers [68] were used also in the works [1, 9], [10], [11], [12] where the eigenvalue problem (1.4), (1.10) was studied in spherical coordinates in the ball using the Bessell functions.

Cantarella, DeTurck, Gluck, and Teytel presented in Table 1 of paper [5] results of their calculations (using Bessel’s functions) of the minimal eigenvalues λ 1 ( 1 ) for different values of R 1 and R 2.

In this paper, we calculate for the first time in the literature the products of the correspondent numbers λ 1 ( 1 ) and R 1R 2 from Table 1 of [5]. The results of our calculations are presented in the right column of Figure 1.

Figure 1:

Calculation of dimensionless eigenvalues x 1.

Reading the right column of Figure 1 down from the top to the bottom we see that numbers x 1 = λ 1 ( 1 ) ( R 1 R 2 ) monotonously decrease and tend to the number of π = 3.1415926535…. Therefore our calculations prove that the dimensionless eigenvalues x 1 = λ 1 ( 1 ) ( R 1 R 2 ) converge to the number of π when (R 1R 2) → 0 (hence the dimensionless parameter μ = R 1 R 2/(R 1R 2) → ∞). As a consequence of this we get the asymptotics

(1.11) λ 1 ( 1 ) π R 1 R 2 , as μ .

This result is the first appearance of the scale invariance of the spectral problem (1.4), (1.10). In Section 3, we prove asymptotics (1.11) more rigorously.

Note that for R 1 = 10 and R 2 = 9.996665555, the corresponding value of parameter μ is μ = 8996999.5. For R 1 = 100 and R 2 = 99.99996667 (in the bottom line of Figure 1), the value of μ is enormous: μ = 9001797 × 106.

Due to the scale invariance (1.5) of the Beltrami Eq. (1.4) the eigenvalue problem (1.4), (1.10) is invariant with respect to the scale transformations

(1.12) R 1 c R 1 , R 2 c R 2 , λ 1 c λ ,

where c > 0 is an arbitrary parameter. We define the dimensionless parameters

(1.13) x = λ ( R 1 R 2 ) > 0 , μ = R 1 R 2 ( R 1 R 2 ) 2 > 0 ,

that are invariant under the group of scaling transformations (1.12). We will call parameter x the dimensionless eigenvalue.

We construct in this paper two infinite series of eigenvalues λ n and λ ̃ m and the corresponding eigenfields V n (x, A) and V ̃ m ( x , A , B ) which depend on arbitrary constant vectors A , B R 3 . Each eigenvalue λ n has multiplicity 3 and each eigenvalue λ ̃ m has multiplicity 5. We study the eigenvalue problem (1.4), (1.10) in an arbitrary shell S R 2 R 1 (1.9) and use only Cartesian coordinates x 1, x 2, x 3 and elementary functions of them. In Section 3, we derive the first series of eigenvalues λ n = x n /(R 1R 2) where x n are roots of equation

(1.14) tan ( x ) = x 1 + μ x 2 .

In Section 3, we show that any root x n of Eq. (1.14) defines the eigenvalue λ n :

(1.15) λ n = 4 μ + 1 + 1 2 R 1 x n

of the spectral problem (1.4) and (1.10). Here, radius R 1 is arbitrary due to the scale invariance (1.12). Radius R 2 is connected with R 1 by relation

(1.16) R 2 = 4 μ + 1 1 4 μ + 1 + 1 R 1 .

In Section 4, we derive the second series of eigenvalues λ ̃ m = x ̃ m / ( R 1 R 2 ) where x ̃ m are the roots of equation

(1.17) tan ( x ) = x + μ x 3 / 3 1 + ( μ 1 ) x 2 / 3 + μ 2 x 4 / 9 .

We prove that any root x ̃ m of Eq. (1.17) defines the eigenvalue λ ̃ m :

(1.18) λ ̃ m = 4 μ + 1 + 1 2 R 1 x ̃ m ,

where radius R 1 is arbitrary in accordance with the scale invariance (1.12) and radius R 2 is defined by Eq. (1.16).

## Remark 1

Cantarella, DeTurck, Gluck, and Teytel wrote on p. 2767 of [5]:

“Theorem B. For the spherical shell B 3(a, b) of inner radius a and outer radius b, the eigenvalue of curl having least absolute value is λ 1 ( 1 ) , where λ 1 ( 1 ) is the smallest of the infinite sequence of positive numbers x k that satisfy

J 3 / 2 ( a x ) Y 3 / 2 ( b x ) J 3 / 2 ( b x ) Y 3 / 2 ( a x ) = 0 .

Thus the eigenvalues λ 1 ( 1 ) are defined in paper [5] by this equation in terms of Bessel’s functions. The scale invariance (1.12) and the dimensionless Eqs. (1.14) and (1.17) were not derived in paper [5].

Let us prove that for any given radii R 1 and R 2 no one eigenvalue λ n coincides with an eigenvalue λ ̃ m . Indeed, let λ be the common value of λ n = λ ̃ m . Then the dimensionless eigenvalue x = λ(R 1R 2) satisfies both Eqs (1.14) and (1.17). Subtracting Eqs (1.14) and (1.17) and reducing to the common denominator we arrive at equation

(1.19) 2 μ 2 x 2 + 9 μ + 3 = 0

that has no real solutions. Hence there are no real numbers x satisfying to both Eqs (1.14) and (1.17). Therefore for any R 1 and R 2 the eigenvalues λ n and λ ̃ m are not equal for any n and m.

In Section 3, we prove that all eigenvalues λ k for R 2R 1 have the following asymptotics

λ k k π R 1 R 2 , as μ .

This asymptotics for k = 1 reduces to the special asymptotics (1.11). Equation (1.14) with arbitrary values of parameter μ > 0 (that means for arbitrary R 1 and R 2) yields the following asymptotics for eigenvalues λ n :

λ n n π R 1 R 2 , as n .

In Section 4, we prove the following asymptotics for the eigenvalues λ ̃ m :

λ ̃ k k π R 1 R 2 as μ , λ ̃ m m π R 1 R 2 as m .

In Section 3, we present the three-dimensional (3D) space of eigenfields V n (x, A) in explicit form that yields that any linear combination of the axisymmetric eigenfields V n (x, A) with arbitrary axes of symmetry is again axisymmetric.

We present in Section 5 in an explicit form the linear five-dimensional (5D) space of eigenfields V ̃ m ( x , A , B ) and prove in Section 6 that the eigenfields V ̃ m ( x , A , B ) with noncollinear constant vectors A, B are not axisymmetric. The corresponding dynamical systems d x / d t = V ̃ m ( x , A , B ) do not have first integrals. A recent numerical investigation [4] proves that some of these systems possess chaotic streamlines and therefore are not integrable.

The special eigenfields κ V ̃ m ( x , A , A ) with B = κ A are axisymmetric. The general linear combinations of the axisymmetric eigenfields V ̃ m ( x , A k , A k ) are not axisymmetric. In Section 7, we demonstrate that topologies of the axisymmetric fluid flows V ̃ m ( x , A , A ) and V n (x, A) are completely different. All eigenfields V ̃ m ( x , A , B ) are presented in explicit form in terms of elementary functions of the Cartesian coordinates x 1, x 2, x 3.

## 2 Functions H n (λR) and their properties

Using the method of Chandrasekhar and Kendall [8], we study vector fields

V ( x ) = curl U ( x ) + 1 λ curl curl U ( x ) ,

where U(x) satisfies the vector Helmholtz equation

(2.1) Δ U ( x ) = λ 2 U ( x ) .

In view of the identity curl curl U(x) = grad(div U(x)) − ΔU(x) and Eq. (2.1), we get

(2.2) V ( x ) = curl U ( x ) + λ U ( x ) + 1 λ grad ( d i v U ( x ) ) .

Applying to V(x) the operator curl and using the identity curl(grad F(x)) = 0 we find that vector field V(x) (2.2) satisfies the Beltrami Eq. (1.4).

We consider vector field U(x) (2.1) of the form U(x) = H 1(λR)A, where A is an arbitrary vector and function H 1(λR) satisfies the spherically symmetric Helmholtz equation

(2.3) Δ u = d 2 u ( R ) d R 2 + 2 R d u ( R ) d R = λ 2 u ( R ) ,

variable R is the spherical radius R = x 1 2 + x 2 2 + x 3 2 . All solutions to the Helmholtz Eq. (2.3) have the form

(2.4) H 1 ( λ R ) = H 1 ( v ) = a sin ( v ) v + b cos ( v ) v , H 1 ( v ) = 1 v d H 0 ( v ) d v ,

where v = λR, and a, b are arbitrary constants.[1] Here, function H 0(v) is

(2.5) H 0 ( λ R ) = H 0 ( v ) = a cos v + b sin v .

In this paper, we will use elementary functions

(2.6) H 2 ( v ) = 1 v d H 1 ( v ) d v = a v 2 cos v sin v v b v 2 sin v + cos v v ,

(2.7) H 3 ( v ) = 1 v d H 2 ( v ) d v = a v 4 ( 3 v 2 ) sin v v 3 cos v + b v 4 3 sin v + ( 3 v 2 ) cos v v ,

H 4 ( v ) = 1 v d H 3 ( v ) d v = a v 6 ( 6 v 2 15 ) sin v v ( v 2 15 ) cos v + b v 6 ( v 2 15 ) sin v + ( 6 v 2 15 ) cos v v .

Let us define by induction the functions

(2.8) H n + 1 ( v ) = 1 v d H n ( v ) d v .

## Lemma 1

Functions H 0(v), H 1(v), …, H n (v), … with arbitrary constants a and b satisfy an infinite series of identities

(2.9) H n ( v ) + ( 2 n + 1 ) H n + 1 ( v ) + v 2 H n + 2 ( v ) = 0 .

## Proof

It is easy to verify using formulas (2.4), (2.5), and (2.6) the identity (2.9) for n = 0:

H 0 ( v ) + H 1 ( v ) + v 2 H 2 ( v ) = 0 .

Assume that identity (2.9) is true for an integer n ≥ 0 and apply to it the operator v −1d/dv. Then using the definitions (2.8) we get

H n + 1 ( v ) + ( 2 n + 1 ) H n + 2 ( v ) + 2 H n + 2 ( v ) + v 2 H n + 3 ( v ) = 0 ,

that is the identity (2.9) for n + 1. Hence identities (2.9) are proven by induction for all n ≥ 0. □

We apply identity (2.9) for n = 1 and n = 2 in Sections 3, 4, 5, and 7 of this paper.

## 3 The first series of eigenvalues λ n and axisymmetric eigenfields V n (x, A)

Substituting formula U(x) = H 1(v)A into Eq. (2.2) we derive

(3.1) V ( x , A ) = grad H 1 ( v ) × A + λ H 1 ( v ) A + 1 λ grad grad H 1 ( v ) A = λ 2 1 v d H 1 ( v ) d v x × A + λ H 1 ( v ) + 1 v d H 1 ( v ) d v A + λ 3 1 v d d v 1 v d H 1 ( v ) d v ( x A ) x .

## Remark 2

Vector field (3.1) evidently is invariant with respect to rotations around the axis having direction A. Vector fields (3.1) for any vector A and for λ = 1 and H 1(v) = (sin R)/R (that corresponds to b = 0) are equivalent to the well-known in plasma physics spheromak exact solution [1, 7, 13, 14].

Using formulas (2.6) and (2.7) we represent vector field V(x, A) (3.1) in the form

(3.2) V ( x , A ) = λ 2 H 2 ( v ) x × A + λ H 1 ( v ) + H 2 ( v ) A + λ 3 H 3 ( v ) ( x A ) x .

Hence we get

(3.3) x V ( x , A ) = λ H 1 ( v ) + H 2 ( v ) ( x A ) + λ 3 H 3 ( v ) ( x x ) ( x A ) = λ H 1 ( v ) + H 2 ( v ) + v 2 H 3 ( v ) ( x A ) ,

where we put λ 2(xx) = λ 2 R 2 = v 2. Identity (2.9) for n = 1 has the form

(3.4) H 1 ( v ) + 3 H 2 ( v ) + v 2 H 3 ( v ) = 0 .

In view of identity (3.4), Eq. (3.3) becomes

(3.5) x V ( x , A ) = 2 λ H 2 ( v ) ( x A ) .

Therefore the two nonpenetration boundary conditions (1.10) are satisfied for the fluid flow V(x, A) (3.2) if and only if

(3.6) H 2 ( v 1 ) = 0 , H 2 ( v 2 ) = 0 , v 1 = λ R 1 , v 2 = λ R 2 .

Multiplying Eq. (2.6) with v 2/(a cos v) we find

v 2 a cos v H 2 ( v ) = 1 tan v v b a 1 v + tan v .

Hence equation H 2(v) = 0 takes the form

(3.7) b a = v tan v 1 + v tan v .

Therefore Eqs (3.6) and (3.7) yield

(3.8) b a = v 1 tan v 1 1 + v 1 tan v 1 = v 2 tan v 2 1 + v 2 tan v 2 .

The second equality implies

v 1 + v 1 v 2 tan v 2 tan v 1 v 2 tan v 1 tan v 2 = v 2 + v 1 v 2 tan v 1 tan v 2 v 1 tan v 1 tan v 2 .

Collecting the similar terms we get

( v 1 v 2 ) ( 1 + tan v 1 tan v 2 ) = ( 1 + v 1 v 2 ) ( tan v 1 tan v 2 ) .

Hence we find

(3.9) tan v 1 tan v 2 1 + tan v 1 tan v 2 = v 1 v 2 1 + v 1 v 2 .

Substituting v 1 = λR 1 and v 2 = λR 2 into Eq. (3.9) and using the trigonometric identity

(3.10) tan α tan β 1 + tan α tan β = tan ( α β )

we get from Eq. (3.9) the equation for the eigenvalues λ:

(3.11) tan λ ( R 1 R 2 ) = λ ( R 1 R 2 ) 1 + λ 2 R 1 R 2 .

Substituting v 1 = λR 1 and v 2 = λR 2 into Eq. (3.8) we get

(3.12) b a = λ R 1 tan ( λ R 1 ) 1 + λ R 1 tan ( λ R 1 ) = λ R 2 tan ( λ R 2 ) 1 + λ R 2 tan ( λ R 2 ) .

Hence function H 1(λR) (2.4) has the form

(3.13) H 1 ( λ R ) = a sin ( λ R ) λ R + λ R 1 tan ( λ R 1 ) 1 + λ R 1 tan λ R 1 cos ( λ R ) λ R ,

where a is an arbitrary parameter.

Let us express the dimensionless quantities λR 1 and λR 2 in terms of variables (1.13). We have

(3.14) λ R 1 = α 1 x , α 1 = R 1 R 1 R 2 , λ R 2 = α 2 x , α 2 = R 2 R 1 R 2 .

Evidently α 1α 2 = 1 and α 1 α 2μ. Hence α 1 satisfies equation α 1 2 α 1 μ = 0 . Therefore by Vieta formula we find

(3.15) α 1 = 1 2 4 μ + 1 + 1 , α 2 = 1 2 4 μ + 1 1 .

Equations (3.14) and (3.15) yield

(3.16) λ = α 1 x R 1 = 4 μ + 1 + 1 2 R 1 x .

Another consequence of Eqs (3.14) and (3.15) is

(3.17) λ R 1 = x 2 4 μ + 1 + 1 , λ R 2 = x 2 4 μ + 1 1 .

Substituting formulas (3.17) into Eq. (3.13) we get

(3.18) H 1 ( λ R ) = a sin ( λ R ) λ R + α 1 x tan ( α 1 x ) 1 + α 1 x tan ( α 1 x ) cos ( λ R ) λ R .

Equation (3.11) evidently is invariant with respect to the group of scale transformations (1.12). In the dimensionless variables x = λ(R 1R 2) and μ = R 1 R 2 / ( R 1 R 2 ) 2 (1.13) Eq. (3.11) takes the form

(3.19) tan ( x ) = x 1 + μ x 2 .

Any root x n of Eq. (3.19) defines the eigenvalue λ n according to formula (3.16):

(3.20) λ n = 4 μ + 1 + 1 2 R 1 x n .

Here, radius R 1 is arbitrary in accordance with the scale invariance (1.12). Equation (3.17) implies that radius R 2 is connected with R 1 by relation

(3.21) R 2 = 4 μ + 1 1 4 μ + 1 + 1 R 1 .

Formula (3.21) agrees with the definition (1.13) of parameter μ = R 1 R 2 / ( R 1 R 2 ) 2 .

Since function tan(x) is π-periodic and tan(x) → ±∞ as x → (n ± 1/2)π we see that Eq. (3.19) for any constant μ > 0 has infinitely many roots x n that satisfy the inequalities n π < x n < n + 1 / 2 π . The maximal value of function f(x) = x/(1 + μx 2) is 1 / ( 2 μ ) that is attained at x * = 1 / μ . Therefore at μ ≫ 1 function f ( x ) 1 / ( 2 μ ) 1 for all x. Hence solutions x k of Eq. (3.19) are close to the roots of function tan(x) that means to . Hence we get the asymptotics at μ ≫ 1 or at R 2R 1:

(3.22) λ k k π R 1 R 2 , as R 2 R 1 .

Since the right hand side of Eq. (3.19) tends to zero as x → ∞ we get that Eq. (3.19) with arbitrary value of parameter μ > 0 yields the following asymptotics for eigenvalues λ n as n → ∞:

(3.23) λ n n π R 1 R 2 .

## Example 1

Let R 1/R 2 = 10. Then parameter μ = 10/81 = 0.1234568. For any root x k to Eq. (3.19) the corresponding eigenvalue λ k = x k /(R 1R 2) = x k /(9R 2). The first five roots of Eq. (3.19) are:

(3.24) x 1 = 4.070059 , x 2 = 7.061906 , x 3 = 10.065190 , x 4 = 13.099750 , x 5 = 16.160470 .

## Example 2

Let R 1/R 2 = 2. Then parameter μ = 2 and any root x k to Eq. (3.19) defines an eigenvalue λ k = x k /(R 1R 2) = x k /R 2. The first five roots to Eq. (3.19) are:

(3.25) x 1 = 3.286005 , x 2 = 6.360678 , x 3 = 9.477196 , x 4 = 12.605920 , x 5 = 15.739670 .

Comparing these sequences with the sequence of numbers z k = :

(3.26) z 1 = 3.141593 , z 2 = 6.283185 , z 3 = 9.424778 , z 4 = 12.56637 , z 5 = 15.70796 ,

we observe that the roots x k (3.25) corresponding to μ 2 = 2 tend to the sequence z k = faster than the roots (3.24) corresponding to μ 1 = 0.1234568.

Let x n be a root of Eq. (3.19). Substituting x = x n into formula (3.18) we obtain function

(3.27) H 1 . n ( λ n R ) = a sin ( λ n R ) λ n R + α 1 x n tan ( α 1 x n ) 1 + α 1 x n tan ( α 1 x n ) cos ( λ n R ) λ n R ,

corresponding to the eigenvalue λ n = x n /(R 1R 2). Vector field V n (x, A) (3.1) and (3.2) corresponding to function H 1.n (λ n R) (3.27) satisfies the Beltrami Eq. (1.4) with λ = λ n and the two nonpenetration boundary conditions (1.10). Therefore it is one of the corresponding eigenfields.

## Remark 3

Any linear combination of eigenfields V n x , A k (3.2) depending on arbitrary vectors A k has the form

(3.28) k = 1 N c k V n x , A k = V n x , k = 1 N c k A k ,

due to the linearity of formula (3.2) with respect to vector A. Hence an arbitrary linear combination (3.28) is another Beltrami vector field (3.2). Therefore for any eigenvalue λ n the linear space of the corresponding eigenfields (3.2) has dimension 3. Hence each eigenvalue λ n = x n /(R 1R 2) where x n is a root of Eq. (3.19) has multiplicity 3.

In view of formula (3.2), the eigenfield (3.28) is invariant under rotations around vector k = 1 N c k A k . Formula (3.28) provides an explanation why any linear combination of the axisymmetric eigenfields V n x , A k is also axisymmetric. For the second series of eigenfields in Section 4, this is not so.

## 4 The second series of eigenvalues λ ̃ m and eigenfields V ̃ m ( x , A , B )

For any λ vector field V(x, A) (3.1), (3.2) satisfies Beltrami Eq. (1.4). Hence its derivative ∇ B V(x, A) in direction of any vector B also satisfies Eq. (1.4). By definition we have

(4.1) B V ( x , A ) = i = 1 3 V ( x , A ) x i B i .

For functions H k (v) we find using the equality v = λR and definition (2.8):

(4.2) B H k ( v ) = i = 1 3 d H k ( v ) d v v x i B i = d H k ( v ) d v λ R ( x B ) = λ 2 v d H k ( v ) d v ( x B ) = λ 2 H k + 1 ( v ) ( x B ) .

Applying formulas (4.1) and (4.2) to vector field (3.2) we find

(4.3) B V ( x , A ) = λ 2 H 2 ( v ) B × A + λ 4 H 3 ( v ) ( x B ) x × A + λ 3 H 2 ( v ) + H 3 ( v ) ( x B ) A + λ 5 H 4 ( v ) ( x A ) ( x B ) x + λ 3 H 3 ( v ) ( A B ) x + ( x A ) B .

Permuting here vectors A and B we get vector field

(4.4) A V ( x , B ) = λ 2 H 2 ( v ) A × B + λ 4 H 3 ( v ) ( x A ) x × B + λ 3 H 2 ( v ) + H 3 ( v ) ( x A ) B + λ 5 H 4 ( v ) ( x B ) ( x A ) x + λ 3 H 3 ( v ) ( B A ) x + ( x B ) A .

Adding formulas (4.3) and (4.4), we get Beltrami field that is symmetric with respect to vectors A and B:

(4.5) V ̃ ( x , A , B ) = 1 2 λ 3 A V ( x , B ) + B V ( x , A ) = 1 2 λ H 3 ( v ) ( x B ) x × A + ( x A ) x × B + 1 2 H 2 ( v ) + H 3 ( v ) ( x A ) B + ( x B ) A + λ 2 H 4 ( v ) ( x A ) ( x B ) x + 1 2 H 3 ( v ) 2 ( A B ) x + ( x B ) A + ( x A ) B .

Formula (4.5) and λ 2(xx) = λ 2 R 2 = v 2 yield

(4.6) x V ̃ ( x , A , B ) = H 2 ( v ) + H 3 ( v ) ( x A ) ( x B ) + v 2 H 4 ( v ) ( x A ) ( x B ) + ( x x ) H 3 ( v ) ( A B ) + H 3 ( v ) ( x A ) ( x B ) = H 3 ( v ) ( A B ) ( x x ) + H 2 ( v ) + 2 H 3 ( v ) + v 2 H 4 ( v ) ( x A ) ( x B ) .

Identity (2.9) for n = 2 takes the form

(4.7) H 2 ( v ) + 5 H 3 ( v ) + v 2 H 4 ( v ) = 0 .

Equation (4.6) in view of identity (4.7) becomes

(4.8) x V ̃ ( x , A , B ) = H 3 ( v ) ( A B ) ( x x ) 3 ( x A ) ( x B ) .

## Remark 4

Equation (4.8) implies that on each sphere S * 2 satisfying equation R = R *, where H 3(v *) = H 3(λR *) = 0 we have x V ̃ ( x , A , B ) = 0 . That means vector field V ̃ ( x , A , B ) is tangent to the spheres S * 2 : R = R *. Hence the spheres S * 2 are invariant submanifolds for the flows V ̃ ( x , A , B ) (4.5) with arbitrary vectors A and B. Therefore the two nonpenetration boundary conditions (1.10) are satisfied for the fluid flows V ̃ ( x , A , B ) (4.5) if and only if

(4.9) H 3 ( v 1 ) = 0 , H 3 ( v 2 ) = 0 , v 1 = λ R 1 , v 2 = λ R 2 .

Multiplying function H 3(v) (2.7) with v 4/(a cos v) we get

v 4 a cos v H 3 ( v ) = ( 3 v 2 ) tan v v 3 + b a 3 tan v + 3 v 2 v .

Hence equation H 3(v) = 0 is equivalent to

(4.10) b a = 3 v ( 3 v 2 ) tan v 3 v tan v + 3 v 2 .

Since b/a is const, the two nonpenetration conditions (4.9) and Eq. (4.10) imply the equality

(4.11) b a = 3 v 1 ( 3 v 1 2 ) tan v 1 3 v 1 tan v 1 + 3 v 1 2 = 3 v 2 ( 3 v 2 2 ) tan v 2 3 v 2 tan v 2 + 3 v 2 2 ,

where v 1 = λR 1 and v 2 = λR 2. Equation (4.11) leads to

9 v 1 v 2 tan v 2 + 3 ( 3 v 2 2 ) v 1 3 v 2 ( 3 v 1 2 ) tan v 1 tan v 2 ( 3 v 1 2 ) ( 3 v 2 2 ) tan v 1 = 9 v 1 v 2 tan v 1 + 3 ( 3 v 1 2 ) v 2 3 v 1 ( 3 v 2 2 ) tan v 1 tan v 2 ( 3 v 1 2 ) ( 3 v 2 2 ) tan v 2 .

Collecting here similar terms we find

3 ( v 1 v 2 ) ( 3 + v 1 v 2 ) 1 + tan v 1 tan v 2 = 9 v 1 v 2 + ( 3 v 1 2 ) ( 3 v 2 2 ) ( tan v 1 tan v 2 ) .

This equation evidently implies

tan v 1 tan v 2 1 + tan v 1 tan v 2 = 3 ( v 1 v 2 ) ( 3 + v 1 v 2 ) 9 v 1 v 2 + ( 3 v 1 2 ) ( 3 v 2 2 ) .

Applying here the trigonometric identity (3.10) and substituting v 1 = λR 1 and v 2 = λR 2 we get the equation for the eigenvalues λ:

(4.12) tan λ ( R 1 R 2 ) = 3 λ ( R 1 R 2 ) ( 3 + λ 2 R 1 R 2 ) 9 λ 2 R 1 R 2 + 9 3 λ 2 ( R 1 2 + R 2 2 ) + λ 4 R 1 2 R 2 2 .

Equation (4.12) evidently is invariant with respect to the scale transformations (1.12). Using formulas (3.17) it is easy to verify that in the dimensionless variables xμ (1.13) Eq. (4.12) takes the form

(4.13) tan ( x ) = x + μ x 3 / 3 1 + ( μ 1 ) x 2 / 3 + μ 2 x 4 / 9 .

Due to Eq. (3.16), any root x ̃ m of Eq. (4.13) defines the eigenvalue λ ̃ m :

(4.14) λ ̃ m = α 1 x ̃ m R 1 = 4 μ + 1 + 1 2 R 1 x ̃ m .

Here, radius R 1 is arbitrary in agreement with the scale invariance (1.12). For any fixed μ (1.13) radius R 2 is connected with R 1 by Eq. (1.16).

## Example 3

Let R 1 / R 2 = ( 3 + 5 ) / 2 = 2.618034 . Then μ = 1 and any root x ̃ m to Eq. (4.13) defines an eigenvalue λ ̃ m = x ̃ m / ( R 1 R 2 ) = 2 x ̃ m / [ ( 1 + 5 ) R 2 ] = x ̃ m / ( 1.618034 R 2 ) . The first five roots to Eq. (4.13) with μ = 1 are:

(4.15) x ̃ 1 = 3.871221 , x ̃ 2 = 6.725464 , x ̃ 3 = 9.732386 , x ̃ 4 = 12.800570 , x ̃ 5 = 15.896620 .

It is evident that sequence (4.15) is an approximation of sequence z k = (3.26).

Since tan(x) is periodic with period π, tan() = 0 and tan(x) → ±∞ as x → (k ± 1/2)π we get that Eq. (4.13) has infinitely many roots. The function f 2(x) in the right hand side of Eq. (4.13) tends to zero when either μ → ∞ (that means R 2R 1) or when x → ∞. Hence we get the asymptotics

(4.16) λ ̃ k k π R 1 R 2 as R 2 R 1 , λ ̃ m m π R 1 R 2 as m .

Let λ ̃ m be a root of Eq. (4.12). Then from Eq. (4.14), we get λ ̃ m R 1 = α 1 x ̃ m where x ̃ m satisfies Eq. (4.13) Substituting v 1 = λ ̃ m R 1 = α 1 x ̃ m into the first Eq. (4.11) we find the constant

(4.17) b a = 3 α 1 x ̃ m ( 3 ( α 1 x ̃ m ) 2 ) tan ( α 1 x ̃ m ) 3 α 1 x ̃ m tan ( α 1 x ̃ m ) + 3 ( α 1 x ̃ m ) 2 .

Therefore the corresponding function H 1(λR) (2.4) takes the form

(4.18) H 1 . m ( v ) = a sin ( v ) v + 3 α 1 x ̃ m ( 3 ( α 1 x ̃ m ) 2 ) tan ( α 1 x ̃ m ) 3 α 1 x ̃ m tan ( α 1 x ̃ m ) + 3 ( α 1 x ̃ m ) 2 cos ( v ) v ,

where a is an arbitrary constant. Vector fields V ̃ m ( x , A , B ) (4.5) with λ = λ ̃ m and constant b/a (4.17) and corresponding to the function H 1.m (v) (4.18) satisfy the Beltrami Eq. (1.4) with λ = λ ̃ m in the spherical shell S R 2 R 1 (1.9) and the two nonpenetration boundary conditions (1.10) on the spheres R = R 1 and R = R 2.

Vector fields V ̃ ( x , A , B ) (4.5) are bilinear symmetric functions of vectors A and B because of equality V ̃ ( x , A , B ) = V ̃ ( x , B , A ) . Expanding vectors A and B in an orthonormal basis e ̂ 1 , e ̂ 2 , and e ̂ 3 we find that any eigenfield V ̃ m ( x , A , B ) is a linear combination of six eigenfields V ̃ m ( x , e ̂ i , e ̂ j ) = V ̃ m ( x , e ̂ j , e ̂ i ) . However, in view of identity (5.1) proved in Section 5 below only five of these eigenfields are linearly independent. Therefore for any eigenvalue λ ̃ m the space of eigenfields V ̃ m ( x , A , B ) has dimension 5. Hence the multiplicity of each eigenvalue λ ̃ m is equal to 5.

## 5 An identity for axisymmetric eigenfields

Vector field V ̃ ( x , A , A ) (4.5) with B = A has the form

(5.1) V ̃ ( x , A , A ) = λ H 3 ( v ) ( x A ) x × A + λ 2 H 4 ( v ) ( x A ) 2 x + H 2 ( v ) + H 3 ( v ) ( x A ) A + H 3 ( v ) ( A A ) x + ( x A ) A .

Vector field V ̃ ( x , A , A ) (5.1) evidently is invariant under rotations around the axis generated by vector A. The corresponding fluid flow is integrable and its streamlines belong to certain axisymmetric tori T 2 , see Section 7.

## Lemma 2

Let vectors A 1, A 2, and A 3 are mutually orthogonal: (A i A j ) = 0 and ij. Then the identity holds:

(5.2) 1 | A 1 | 2 V ̃ ( x , A 1 , A 1 ) + 1 | A 2 | 2 V ̃ ( x , A 2 , A 2 ) + 1 | A 3 | 2 V ̃ ( x , A 3 , A 3 ) = 0 .

## Proof

Vector fields e ̂ k = A k / | A k | have unit lengths and are mutually orthogonal. Evidently we have V ̃ ( x , A k , A k ) / | A k | 2 = V ̃ ( x , e ̂ k , e ̂ k ) . Vector field V ̃ ( x , e ̂ k , e ̂ k ) (4.5) has the form

(5.3) V ̃ ( x , e ̂ k , e ̂ k ) = λ H 3 ( v ) x × ( ( x e ̂ k ) e ̂ k ) + H 2 ( v ) + H 3 ( v ) ( x e ̂ k ) e ̂ k + λ 2 H 4 ( v ) ( x e ̂ k ) 2 x + H 3 ( v ) × ( e ̂ k e ̂ k ) x + ( x e ̂ k ) e ̂ k .

The following identities evidently hold in the basis e ̂ 1 , e ̂ 2 , and e ̂ 3 :

k = 1 3 ( x e ̂ k ) e ̂ k = x , k = 1 3 ( x e ̂ k ) 2 = | x | 2 = R 2 , k = 1 3 ( e ̂ k e ̂ k ) = 3 .

Hence using Eq. (5.3) and equation λ 2 R 2 = v 2 we find

k = 1 3 V ̃ ( x , e ̂ k , e ̂ k ) = λ H 3 ( v ) [ x × x ] + H 2 ( v ) + 5 H 3 ( v ) + v 2 H 4 ( v ) x .

This expression vanishes due to the identity x × x = 0 and identity (4.7). Hence we get

k = 1 3 1 | A k | 2 V ̃ ( x , A k , A k ) = k = 1 3 V ̃ ( x , e ̂ k , e ̂ k ) = 0 ,

that proves the identity (5.2). □

## Remark 5

An arbitrary linear combination of two axisymmetric eigenfields V ̃ m ( x , A 1 , A 1 ) and V ̃ m ( x , A 2 , A 2 ) is not axisymmetric. However, if vectors A 1 and A 2 are orthogonal and have equal lengths A = |A 1| = |A 2| then using identity (5.2) of Lemma 2 we get

(5.4) V ̃ m ( x , A 1 , A 1 ) + V ̃ m ( x , A 2 , A 2 ) = V ̃ m ( x , A 3 , A 3 ) ,

where A 3 = (A 1 × A 2)/A, |A 3| = A. Equations (5.4), (5.1) imply that for this case the eigenfield V ̃ m ( x , A 1 , A 1 ) + V ̃ m ( x , A 2 , A 2 ) is axisymmetric with respect to the axis generated by vector A 3 that is orthogonal to both vectors A 1 and A 2.

## 6 Eigenfields V ̃ m ( x , A , B ) with any noncollinear vectors A, B are not axisymmetric

Fluid streamlines defined by an eigenfield V ̃ m ( x , A , B ) are trajectories of the dynamical system

(6.1) d x ( t ) d t = V ̃ m ( x ( t ) , A , B ) .

## Theorem 1

For any noncollinear vectors A and B vector field V ̃ m ( x , A , B ) and dynamical system (6.1) are not axisymmetric.

## Proof

On the boundaries of the spherical shell (1.9) S k 2 : R = R k and v = v k = λR k we have H 3(v k ) = 0. Hence from identity (4.7) we get v k 2 H 4 ( v k ) = H 2 ( v k ) . Note that H 2(v k ) ≠ 0 because if both H 3(v k ) = 0 and H 2(v k ) = 0 then from identities (2.9) for n = 1 and n = 0 we get H 1(v k ) = 0 and H 0(v k ) = 0 that implies a 2 + b 2 = 0. Therefore in view of Eq. (4.5) dynamical system (6.1) at H 3(v k ) = 0 takes the form

(6.2) d x d t = 1 2 H 2 ( v k ) ( x A ) B + ( x B ) A 2 R k 2 ( x A ) ( x B ) x .

System (6.2) evidently has critical points

(6.3) x k ± = ± R k | A × B | A × B .

Note that A × B ≠ 0 because vectors A and B are noncollinear. The critical points (6.3) are intersections of the sphere S k 2 with the axis

(6.4) L : x ( t ) = r ( t ) ( A × B ) .

Substituting x = r(A × B) into formula (4.5) and using ((A × B) ⋅ A) = 0 and ((A × B) ⋅ B) = 0 we get

V ̃ ( r ( A × B ) , A , B ) = H 3 ( v ) ( A B ) r ( A × B ) .

Therefore dynamical system (6.1) on the axis L (6.4) reduces to equation

(6.5) d r ( t ) d t = ( A B ) H 3 ( v ( t ) ) r ( t ) ,

where v(t) = λ|r(t)‖A × B|. Equation (6.5) yields that the axis L is an invariant submanifold of system (6.1).

Equation (6.5) shows that if (AB) = 0 then the axis L consists of critical points of the system (6.1). Equation (6.5) defines nontrivial dynamics on the line L if (AB) ≠ 0. In both cases, if system (6.1) is axisymmetric then its axis of symmetry can be only the axis L and then system (6.1) has no isolated critical points outside of the axis L.

However, let us show that system (6.2) has another critical points which belong to the intersection of the sphere S k 2 with the plane P orthogonal to the axis L and have the form

(6.6) x ̃ = α A + β B .

Equation (6.2) at a critical point  x ̃ yields

(6.7) ( x ̃ A ) B + ( x ̃ B ) A = 2 R k 2 ( x ̃ A ) ( x ̃ B ) α A + β B .

Since vectors A and B are linearly independent, we get from Eq. (6.7) two equations

(6.8) 2 α ( x ̃ A ) = R k 2 , 2 β ( x ̃ B ) = R k 2 .

Using expression (6.6) we find

(6.9) α ( x ̃ A ) = α 2 | A | 2 + α β ( A B ) , β ( x ̃ B ) = β 2 | B | 2 + α β ( A B ) ,

(6.10) R k 2 = ( x ̃ x ̃ ) = α 2 | A | 2 + 2 α β ( A B ) + β 2 | B | 2 .

Substituting formulas (6.9) and (6.10) we find that each of the two Eq. (6.8) is equivalent to equation

α 2 | A | 2 = β 2 | B | 2 .

Hence we get β = ±α|A|/|B|. Substituting this into Eq. (6.10) we find

R k 2 = 2 α 2 | A | | B | | A B | ± ( A B ) .

The bracket here is nonzero due to the Cauchy inequality and the noncollinearity of vectors A and B. Therefore we derive

α = ± | B | 2 | A | R k | A B | ± ( A B ) , β = ± | A | 2 | B | R k | A B | ± ( A B ) .

Hence we find the critical points (6.6):

(6.11) x ̃ = R k | A B | ± ( A B ) ± | B | 2 | A | A ± | A | 2 | B | B .

If system (6.1) were axisymmetric then the orbit S ̃ 1 of any critical point  x ̃ under the rotations around the axis of symmetry L would be a continuous set of critical points. Each of the four critical points x ̃ (6.11) belong to the plane P orthogonal to the axis of symmetry L. Hence their one-dimensional orbits S ̃ 1 ( x ̃ ) consisting of critical points also would belong to the plane P. However, we have demonstrated above that the critical set of system (6.1) in the plane P is not continuous but consists only of the four isolated critical points x ̃ (6.11). The obtained contradiction proves that the dynamical system (6.1) and eigenfield V ̃ m ( x , A , B ) are not axisymmetric if vectors A and B are noncollinear. □

## Remark 6

For any collinear vectors A and B, the axis L (6.4) degenerates into one point x = 0 and therefore the above proof is not applicable. A recent numerical investigation [4] proves that some systems (6.1) with noncollinear vectors A and B possess chaotic streamlines and hence are not integrable.

## 7 Topology of axisymmetric eigenfields

The axisymmetry and incompressibility of eigenfields V ̃ m ( x , A , A ) (5.1) lead to the existence of a first integral of the corresponding dynamical systems

(7.1) d x ( t ) d t = V ̃ m ( x ( t ) , A , A ) .

Dynamical system (7.1) possesses a first integral

(7.2) F 2 ( x ) = H 3 ( v ) ( x A ) Z ( x ) , Z ( x ) = ( A A ) ( x x ) ( x A ) 2 .

Indeed, differentiating function F 2(x(t)) along the flow (7.1) we find

(7.3) d F 2 ( x ( t ) ) d t = ( A A ) ( x A ) H 3 ( v ) ( A A ) ( x x ) 3 ( x A ) 2 × H 2 ( v ) + 5 H 3 ( v ) + v 2 H 4 ( v ) .

Applying in Eq. (7.3) the identity (4.7) we obtain dF 2(x(t))/dt ≡ 0. Therefore function F 2(x) (7.2) is the first integral of dynamical system (7.1).

The existence of the first integral F 2(x) = H 3(v)(xA)Z(x) implies that all surfaces of its constant levels F 2(x) = const are invariant submanifolds of dynamical system (7.1). Inside the spherical shell S R 2 R 1 all surfaces of nonzero levels of first integral F 2(x) = c are compact axisymmetric tori T c 2 . Dynamics of the fluid streamlines (7.1) on the tori T c 2 is quasi-periodic. The invariant surface of zero level F 2(x) = 0 consists of the plane (xA) = 0, the straight line x = c A, and infinitely many spheres S m 2 : R = R m where R m are roots of equation H 3(λR m ) = 0.

The axisymmetry and incompressibility of the flow with velocity field V(x, A) (3.1) lead to the existence of a first integral for the fluid streamlines defined by equation

(7.4) d x ( t ) d t = V ( x ( t ) , A ) .

Dynamical system (7.4) has first integral

(7.5) F 1 ( x ) = H 2 ( v ) Z ( x ) , Z ( x ) = ( A A ) ( x x ) ( x A ) 2 .

Indeed, differentiating function F 1(x(t)) along the flow (7.4) we get equation

(7.6) d F 1 ( x ) d t = 2 λ ( x A ) H 2 ( v ) ( A A ) × H 1 ( v ) + 3 H 2 ( v ) + v 2 H 3 ( v ) ,

where we used equation λ 2(xx) = v 2. Applying to the last formula the identity (3.4) we get from Eq. (7.6) dF 1(x(t))/dt ≡ 0. Hence function F 1(x) (7.5) is a first integral of system (7.4).

The existence of the first integral F 1(x) (7.5) yields that all surfaces of its constant level F 1(x) = const are invariant submanifolds of the system (7.4). Inside the spherical shell S R 2 R 1 all nonzero levels of first integral F 1(x) = c are axisymmetric tori T c 2 . Dynamics of the fluid streamlines (7.4) on the tori T c 2 is quasi-periodic. The surface of zero level F 1(x) = H 2(λR)Z(x) = 0 consists of the straight line x = c A and infinitely many spheres S k 2 : R = R k satisfying equation H 2(λR k ) = 0.

## Remark 7

Topology of the axisymmetric fluid flows V ̃ m ( x , A , A ) and that of V n (x, A) are completely different. Indeed, invariant tori T c 2 for V ̃ m ( x , A , A ) belong either to the half space (xA) > 0 or to the half space (xA) < 0 and there is invariant plane (xA) = 0. For fluid flow V n (x, A) each invariant torus T c 2 belongs to the both sides of any plane (xC) = 0 where C is an arbitrary constant vector.

## 8 Summary

We have studied the boundary eigenvalue problem for vector fields V(x) satisfying the Beltrami equation curl V(x) = λ V(x) inside a spherical shell S R 2 R 1 : R 2 < x 1 2 + x 2 2 + x 3 3 < R 1 with nonpenetration boundary conditions xV(x) = 0 on the spheres R = R 2 and R = R 1. We have introduced the scale invariance (1.12) of the eigenvalue problem and have defined the dimensionless parameters x = λ(R 1R 2) and μ = R 1 R 2 / ( R 1 R 2 ) 2 . We have shown that eigenvalues λ are connected with parameters xμ by equation

(8.1) λ = 4 μ + 1 + 1 2 R 1 x ,

where radius R 1 is arbitrary due to the scale invariance. Radius R 2 = R 1 ( 4 μ + 1 1 ) / ( 4 μ + 1 + 1 ) . We have derived equations

(8.2) tan ( x ) = x 1 + μ x 2 ,

(8.3) tan ( x ) = x + μ x 3 / 3 1 + ( μ 1 ) x 2 / 3 + μ 2 x 4 / 9

for the dimensionless parameters xμ. For any μ > 0 the roots x n of Eq. (8.2) define by formula (8.1) eigenvalues λ n . Any root x ̃ m of Eq. (8.3) specifies eigenvalue λ ̃ m by formula (8.1). Using Eqs. (8.2) and (8.3) we have obtained asymptotics of eigenvalues λ n , λ ̃ m at R 2R 1 (that is equivalent to μ → ∞) and at n, m → ∞ and have proved that the eigenvalues λ n and λ ̃ m are not equal for any n and m.

We have derived in elementary functions the corresponding to λ n 3D linear space of eigenfields V n (x, A) and the corresponding to λ ̃ m 5D linear space of eigenfields V ̃ m ( x , A , B ) where A and B are arbitrary constant vectors in R 3 . We have proved that eigenfields V ̃ m ( x , A , B ) with noncollinear vectors A and B are not axisymmetric and eigenfields V ̃ m ( x , A , A ) are axisymmetric. The topologies of fluid flows V n (x, A), V ̃ m ( x , A , A ) , and V ̃ m ( x , A , B ) with A × B ≠ 0 are completely different.

The constructed eigenfields V n (x, A) and V ̃ m ( x , A , B ) define the incompressible fluid and plasma equilibria in the shell S R 2 R 1 and also time-dependent flows (1.6) and (1.7) of viscous magnetic fluid in the shell.

Corresponding author: Oleg Bogoyavlenskij, Departrment of Mathematics and Statistics, Queen’s University, Kingston, ON, K7L 3N6, Canada, E-mail:

1. Author contribution: The author has accepted responsibility for the entire content of this submitted manuscript and approved submission.

2. Research funding: None declared.

3. Conflict of interest statement: The author declares no conflicts of interest regarding this article.

## References

[1] G. Marsh, Force-free Magnetic Fields. Solutions, Topology and Applications, Singapore, World Scientific Publishing Co, 1996.10.1142/2965Search in Google Scholar

[2] D. Biskamp, Nonlinear Magnetohydrodynamics, Cambridge, Cambridge University Press, 1997.Search in Google Scholar

[3] C. Wang, “Exact solutions of the Navier-Stokes equations - the generalized Beltrami flows, review and extension,” Acta Mech., vol. 81, pp. 69–74, 1990. https://doi.org/10.1007/bf01174556.Search in Google Scholar

[4] O. Bogoyavlenskij and R. Kipka, “Chaotic streaklines in new exact solutions to the Navier - Stokes equations,” Phys. Fluids, vol. 33, pp. 1–7, 2021, Art no. 073110. https://doi.org/10.1063/5.0058350.Search in Google Scholar

[5] J. Cantarella, D. DeTurck, H. Gluck, and M. Teytel, “The spectrum of the curl  operator on spherically symmetric domains,” Phys. Plasmas, vol. 7, pp. 2766–2775, 2000. https://doi.org/10.1063/1.874127.Search in Google Scholar

[6] L. Woltjer, “The Crab Nebula,” Bull. Astron. Inst. Netherlands, vol. 14, pp. 39–80, 1958.Search in Google Scholar

[7] S. Chandrasekhar, “On force-free magnetic fields,” Proc. Natl. Acad. Sci., vol. 42, pp. 1–5, 1956. https://doi.org/10.1073/pnas.42.1.1.Search in Google Scholar PubMed PubMed Central

[8] S. Chandrasekhar and P. C. Kendall, “On force-free magnetic fields,” Astrophys. J., vol. 126, pp. 457–460, 1957. https://doi.org/10.1086/146413.Search in Google Scholar

[9] Z. Yoshida and Y. Giga, “Remarks on spectra of operator rot,” Math. Z., vol. 204, pp. 235–245, 1990. https://doi.org/10.1007/bf02570870.Search in Google Scholar

[10] Z. Yoshida, “Discrete eigenstates of plasma described by the Chandrasekhar-Kendall functions,” Prog. Theor. Phys., vol. 86, no. 1, pp. 45–55, 1991. https://doi.org/10.1143/ptp/86.1.45.Search in Google Scholar

[11] G. F. Torres del Castillo, “Eigenfunctions of the curl operator in spherical coordinates,” J. Math. Phys., vol. 35, pp. 499–507, 1994. https://doi.org/10.1063/1.530740.Search in Google Scholar

[12] E. C. Morse, “Eigenfunctions of the curl in cylindrical geometry,” J. Math. Phys., vol. 46, nos 1–13, p. 113511, 2005. https://doi.org/10.1063/1.2118447.Search in Google Scholar

[13] O. Bogoyavlenskij, “Vortex knots for the spheromak fluid flow and their moduli spaces,” J. Math. Anal. Appl., vol. 450, pp. 21–47, 2017. https://doi.org/10.1016/j.jmaa.2016.12.080.Search in Google Scholar

[14] M. N. Rosenbluth and M. N. Bussac, “MHD stability of spheromak,” Nucl. Fusion, vol. 19, pp. 489–498, 1979. https://doi.org/10.1088/0029-5515/19/4/007.Search in Google Scholar

[15] W. Heisenberg, “Lecture on atomic bomb physics,” in Hitler’s Uranium Club. The Secret Recordings at Farm Hall, J. Bernstein, Ed., Woodbury, New York, American Institute of Physics, 1945, pp. 187–209, 1996.Search in Google Scholar