5 Polar geometry: energy orthogonality

In a full sphere geometry, the energy of a vector field over the unit sphere, written in toroidal-poloidal form and then decomposed into spherical harmonics, may be defined as the sum

     ∑  ∫ 1        ∑  ∫ 1        2
E  =        T2αdr +        l(l +-1)Sα-+ (dS-α)2dr
      α  0          α  0     r2        dr

where α indexes the spherical harmonics.

Auto-orthogonality can be extended to the case for which the basis functions are
(i) regular at the origin
(ii) satisfy relevant boundary conditions at r = 1 and are
(iii) orthogonal with respect to the inner product defined for the toroidal or poloidal scalars respectively

∫ 1                 ∫ 1 l(l + 1)χiχj  dχi dχj
   ϕiϕjdr     or        ----r2-----+ -dr-dr-dr
 0                   0

5.1 Toroidal, unconstrained

The toroidal component of a regular vector field that satisfies no particular boundary condition has an orthogonal basis of the form

            (0,l+1∕2)
Ψn  = r(l+1)Pn-1    (2r2 - 1 ).

5.2 Toroidal, T(1) = 0

The toroidal component of a flow subject to nonslip boundary conditions, or the toroidal component of field in contact with an external electrical insulator simply satisfies regularity, T(1) = 0; an orthogonal basis is of the form

      (l+1)      2  (2,l+1∕2)   2
Ψn = r    (1-  r)P n-1    (2r - 1).

5.3 Poloidal, unconstrained

An alternative to the Worland polynomials [4] (which are orthogonal in a different norm) is the orthogonal basis set

          (                                    )
      (l+1)   (0,l+1 ∕2)   2        (0,l+1∕2)   2
Ψn = r     P n-1    (2r - 1) - Pn-2    (2r - 1)

for n 2.

The basis functions asymptotically collapse to a single family of one-sided Jacobi polynomials:

Ψn ~ r(l+1)P (n-1,l+1∕2)(2r2 - 1)

as n →∞.

5.4 Poloidal, S(1) + lS(1) = 0

The poloidal component of a magnetic field in contact with an external electrical insulator satisfies regularity and (ii) S(1) + lS(1) = 0 l 1 is the spherical harmonic degree. It has an orthogonal basis of the form

      (l+1)(    (0,l+1∕2)   2          (0,l+1∕2)   2        )
Ψn = r     c0P n      (2r - 1)+ c1P n-1    (2r  - 1)+ c2

where

c0 = -2n2(l + 1) - n(l + 1)(2l - 1) - l(2l + 1)
c1 = 2(l + 1)n2 + (2l + 3)(l + 1)n + (2l + 1)2
c2 = 4nl + l(2l + 1)

Distinct from most other closed form expressions for the basis functions given in [23], this has a constant term, although scaling only linearly in n, compared to the Jacobi polynomial coefficients that scale quadratically. Note that the coefficients are in the ratio (1,-1,0) in the limit of large n, so that the basis functions asymptotically collapse to a single family of one-sided Jacobi polynomials:

Ψn ~ r(l+1)P (n-1,l+1∕2)(2r2 - 1)

as n →∞. The figures below show some example basis functions and the asymptotic description of the 11th basis function.


PIC

Figure 6: The first 4 basis functions with l = 3.



PIC

Figure 7: The 11th basis function plotted with r(l+1)P11(-1,l+12)(2r2 - 1) with l = 5.


5.5 Poloidal, S(1) = S(1) = 0

The poloidal component of a nonslip flow satisfies regularity and S(1) = S(1) = 0. Although an orthogonal basis can be written

Ψn = r(l+1)(1-  r2)2Qn (r2)

for some polynomial Qn, this has no [obvious] terse representation in terms of Jacobi polynomials. Instead, it may be written

          (                                            )
      (l+1)     (0,l+1∕2)   2          (0,l+1∕2)   2
Ψn = r     c0P n+1    (2r - 1)+ c1P n      (2r  - 1)+ c2

where

c0 = 2n2 + (2l + 3)n
c1 = -2n2 + (-7 - 2l)n - 2l - 5
c2 = 2l + 4n + 5
The coefficients are in the ratio (1,-1,0) in the limit of large n, so that the basis functions asymptotically collapse to a single family of one-sided Jacobi polynomials:
            (-1,l+1∕2)
Ψn ~ r(l+1)P n+1     (2r2 - 1)

as n →∞.


PIC

Figure 8: The first 4 basis functions for l = 3.


5.6 Poloidal, S(1) = 0

A poloidal component of a flow satisfying a non-penetration condition at r = 1 satisfies S(1) = 0 and has an orthogonal basis of the form

                   (1,l+1∕2)
Ψn = r(l+1)(1-  r2)P n-1    (2r2 - 1).


PIC

Figure 9: The first 4 basis functions for l = 3.


5.7 Poloidal, S(1) = 0

The poloidal component of a magnetic field which describes a purely radial field satisfies S(1) = 0, and has an orthogonal basis of the form

          (                                        )
Ψn = r(l+1)  c0P (n0,l+1 ∕2)(2r2 - 1) + c1P (n0,-l+11∕2)(2r2 - 1)

where

c0 = -(2n - 1)(n + l)
c1 = (2n + 1)(n + l + 1)

Note that the coefficients are in the ratio (1,-1) in the limit of large n, so that the basis functions asymptotically collapse to a single family of one-sided Jacobi polynomials:

Ψn ~ r(l+1)P (n-1,l+1∕2)(2r2 - 1)

as n →∞.

5.8 Poloidal, all-space energy, S(1) + lS(1) = 0

In an electrical insulator defined in the region r > 1 an interior magnetic field may be extended as:

              ∑                        ∑            -Yα--
Bint = ∇ × ∇ ×   Sα(r)Yαˆr,    Bext = -    lαSα (1)∇ (rlα+1),

where the boundary condition

dS
---α(1)+ lαSα (1) = 0
 dr

is satisfied. Noting that

∫             ∑
    B2extdV =     --4π---(lα + 1)l2αS α(1)2
 r>1             2lα + 1

it follows that the total energy over all space of the poloidal magnetic field may be defined as the sum

                   [ ∫                                 ]
     ∑  4πlα(lα +-1)   1l(l +-1)S2α   dS-α 2           2
E =       2lα + 1           r2    + ( dr ) dr + lαSα(1)
      α               0

where α indexes the harmonics.

An orthogonal basis that satisfies the orthogonality

∫  1[l(l + 1)Ψn (r)Ψm (r)   dΨn dΨm ]
     ---------2--------+  ---------dr + lΨn (1)Ψm (1) = δnm
  0          r            dr   dr

and the boundary condition S(1) + lS(1) = 0 is of the form

      (l+1 )(              (0,l+1∕2)  2                      (0,l+1∕2)   2   )
Ψn = r      n(2l+2n- 1)Pn      (2r - 1)- (n+1 )(2n+2l+1 )Pn-1   (2r - 1 )

Note that the coefficients are in the ratio (1,-1) in the limit of large n, so that the basis functions asymptotically collapse to a single family of one-sided Jacobi polynomials:

      (l+1) (-1,l+1∕2)   2
Ψn ~ r    P n       (2r - 1)

as n →∞. The figure shows some example basis functions normalised under the radial norm as defined above with l = 3.


PIC

Figure 10: The first 4 basis functions for l = 3.