The different expressions in Eqs. (
A.8) are similar to the modified Bessel differential equation
$$\begin{aligned} x^2f''(x)+x f'(x)-(x^2+n^2)f(x)=0, \end{aligned}$$
(A.11)
which is solved by the modified Bessel functions of the first and second kind. Therefore, the following ansatz is made:
$$\begin{aligned} G_{rz}(r)&=C I_1(\alpha r)+\hat{C}K_1(\alpha r), \end{aligned}$$
(A.12a)
$$\begin{aligned} G_{zr}(r)&=D I_1(\beta r)+\hat{D}K_1(\beta r). \end{aligned}$$
(A.12b)
Here,
\(I_n(x)\) and
\(K_n(x)\) denote the Bessel functions of the first and second kind, respectively. Since
\(K_n(\alpha r)\) is not defined for
\(r=0\), and the solution requires to be defined for all
\(r\in {\overline{\varOmega }}\),
\(\hat{C}=\hat{D}=0\) can be concluded. In order to solve the homogeneous problem, the ansatz function is substituted in Eq. (
A.11) using the linear coordinate transformation
\(x(r)=\alpha r\) with
$$\begin{aligned} \frac{\partial }{\partial r}f(x(r))&=\frac{\partial f(x(r))}{\partial x}\frac{\partial x}{\partial r}\\ \Leftrightarrow \qquad \frac{\partial f(\alpha r)}{\partial r}&=\alpha \frac{\partial f(x)}{\partial x} \\ \Leftrightarrow \qquad \frac{1}{\alpha }f'(\alpha r)&=f'(x) \end{aligned}$$
and analogously
$$\begin{aligned} \frac{1}{\alpha ^2}f''(\alpha r)=f''(x). \end{aligned}$$
The modified Bessel differential equation can thereby be rewritten as
$$\begin{aligned} x^2I_1''(x)+xI'(x)-(x^2+1)I_1(x)&=0 \end{aligned}$$
(A.13a)
$$\begin{aligned} \Leftrightarrow I_1''(x)+x^{-1}I_1'(x)-(1+x^{-2})I_1(x)&=0 \end{aligned}$$
(A.13b)
$$\begin{aligned} \Leftrightarrow \frac{1}{\alpha ^2}I_1''(\alpha r)+\frac{1}{\alpha ^2 r}I_1'(\alpha r)-\left( 1+\frac{1}{\alpha ^2 r^2}\right) I_1(\alpha r)&=0 \end{aligned}$$
(A.13c)
$$\begin{aligned} \Leftrightarrow I_1''(\alpha r)+\frac{1}{r}I_1'(\alpha r)-\left( \alpha ^2+\frac{1}{r^2}\right) I_1(\alpha r)&=0 \end{aligned}$$
(A.13d)
$$\begin{aligned} \Leftrightarrow I_1''(\alpha r)+\frac{1}{r}I_1'(\alpha r)-\left( \alpha _i^2+\frac{1}{r^2}\right) I_1(\alpha r)&=(\alpha ^2-\alpha _i^2)I_1(\alpha r). \end{aligned}$$
(A.13e)
The identity from Eq. (
A.13e) is used to transfer the system of PDEs into a system of linear equations:
$$\begin{aligned} (\alpha ^2-\alpha _1^2)K_1 D I_1(\alpha r)+(\beta ^2-\alpha _2^2)K_2 C I_1(\beta r)&=0, \end{aligned}$$
(A.14a)
$$\begin{aligned} (\alpha ^2-\alpha _3^2)K_3 D I_1(\alpha r)+(\beta ^2-\alpha _4^2)K_4 C I_1(\beta r)&=0. \end{aligned}$$
(A.14b)
The power series representation of the Bessel function of the first kind is given by:
$$\begin{aligned} I_n(z)=\left( \frac{z}{2}\right) ^n\sum \limits _{k=0}^\infty \frac{\left( \frac{z}{2}\right) ^{2k}}{k!(n+k)!} \end{aligned}$$
If this power series is substituted in Eq. (
A.14)), according to the identity theorem for polynomials, the PDE can only be fulfilled for all
\(k\in \mathbb {N}\) if the equalities
$$\begin{aligned} (\alpha ^2-\alpha _1^2)K_1 D \left( \frac{\alpha r}{2}\right) ^{2k+1}+(\beta ^2-\alpha _2^2)K_2 C \left( \frac{\beta r}{2}\right) ^{2k+1}&=0 \end{aligned}$$
(A.15)
$$\begin{aligned} (\alpha ^2-\alpha _3^2)K_3 D \left( \frac{\alpha r}{2}\right) ^{2k+1}+(\beta ^2-\alpha _4^2)K_4 C \left( \frac{\beta r}{2}\right) ^{2k+1}&=0 \end{aligned}$$
(A.16)
hold true. These two equations can only be fulfilled for all
k if
\(\alpha =\beta \). The homogeneous problem can therefore be expressed as following system of linear equations:
$$\begin{aligned} \left( K_1(\alpha ^2-\alpha _1^2)D+K_2(\alpha ^2-\alpha _2^2)C\right) I_1(\alpha r)&=0, \end{aligned}$$
(A.17a)
$$\begin{aligned} \left( K_3(\alpha ^2-\alpha _3^2)D+K_4(\alpha ^2-\alpha _4^2)C\right) I_1(\alpha r)&=0, \end{aligned}$$
(A.17b)
$$\begin{aligned} \Leftrightarrow \underbrace{\begin{pmatrix} (\alpha ^2-\alpha _1^2)K_1 &{} (\alpha ^2-\alpha _2^2)K_2\\ (\alpha ^2-\alpha _3^2)K_3 &{} (\alpha ^2-\alpha _4^2)K_4 \end{pmatrix}}_{{=}{:}{\textbf{A}}} \begin{pmatrix} D\\ C \end{pmatrix}&={\varvec{0}}. \end{aligned}$$
(A.17c)
Since Eq. (
A.17c) is a homogeneous system of linear equations, the rows must depend on each other, which is fulfilled if the determinant of
\({\textbf{A}}\) vanishes. Otherwise, the solution to the linear equation system would be the trivial solution, i.e.,
\(C=D=0\). Hence, the solution to the homogeneous system of differential equations is found by adding the two linear independent solutions
$$\begin{aligned} G_{rz}(r)&=C_1 I_1(\alpha r)+C_2 I_1(\beta r), \end{aligned}$$
(A.18a)
$$\begin{aligned} G_{zr}(r)&=D_1 I_1(\alpha r)+D_2 I_1(\beta r), \end{aligned}$$
(A.18b)
where
\(\alpha \) and
\(\beta \) can be obtained from solving
$$\begin{aligned} \det ( {\textbf{A}})=(\lambda _{1/2}^2-\alpha _1^2)(\lambda _{1/2}^2-\alpha _4^2)K_1K_4-(\lambda _{1/2}^2-\alpha _2^2)(\lambda _{1/2}^2-\alpha _3^2)K_2K_3{\mathop {=}\limits ^{!}}0 \end{aligned}$$
Finally, after plugging Eqs. (
A.18) into Eq. (
A.8a), one obtains the following relations between
\(C_i\) and
\(D_i\):
$$\begin{aligned} f_1~{:}{=}~\frac{D_1}{C_1}=-\frac{(\alpha ^2-\alpha _2^2)}{(\alpha ^2-\alpha _1^2)}\frac{K_2}{K_1},\quad f_2~{:}{=}~\frac{D_2}{C_2}=-\frac{(\beta ^2-\alpha _2^2)}{(\beta ^2-\alpha _1^2)}\frac{K_2}{K_1}. \end{aligned}$$
(A.19)