I can solve the following differential equation without any trouble using the method of Frobenius $$ x^2 y’’ - (2 + 3x) y = 0. $$ When I put the differential equation in Mathematica, it gives me the solutions in terms of modified Bessel functions of order 3 $$ y(x) = A \sqrt{x} I_3\big(2\sqrt{3x}\big) + B \sqrt{x} K_3\big(2\sqrt{3x}\big). $$ I cannot for the life of me see how to put the given ODE into the form of a modified Bessel equation. Can anyone point me in the right direction?

Some Added Information

Generally, the equation $$ \frac{d}{dx}\left(x^a \frac{dy}{dx}\right) + b x^c y = 0, $$ can be transformed to a Bessel equation with solution $$ y(x) = x^{\nu/\alpha} Z_\nu \left(\alpha\sqrt{|b|} x^{1/\alpha}\right), $$ where $Z_\nu$ is any Bessel function solution of the transformed equation, if we choose $$ \alpha = \frac{2}{c-a+2} \quad \text{and} \quad \nu = \frac{1-a}{c-a+2}, $$ Considering the equations for $\alpha$ and $\nu$, I can see from Mathematica’s solution that $\nu = 3$ and $\alpha = 2$, so I can solve them to find $a = -2$ and $c = -3$. Putting those into the ODE above, we get $$ \frac{d}{dx}\left(x^{-2} \frac{dy}{dx}\right) + b x^{-3} y = 0. $$ Expanding this out, I get $$ \frac{1}{x^2}y’’ - \frac{2}{x^3} y’ - \frac{b}{x^3} y = 0 \quad \text{or} \quad x^2 y’’ - 2x y’ - 3xy = 0. $$ This is my ODE if the $y’$ were instead a $y$. :-( So, I am stuck.

  • $\begingroup$ I think you need a transformation first, as the modified vessel equation has solutions in linear combinations of modified vessel functions, not necessarily $\sqrt{x}I_3$ and $\sqrt{x}K_3$. The degree being 3 should also be a hint. $\endgroup$
    – DaveNine
    Dec 15, 2019 at 22:28
  • $\begingroup$ I’ve tried every transformation I can think of and it’s not happening. I will add some more to my question to show what I have tried. $\endgroup$ Dec 15, 2019 at 23:20

2 Answers 2


Your equation is indeed a transformed modified Bessel equation. To see that, you need to take the modified Bessel equation $$ \xi^2 \frac{d^2\eta}{d\xi^2} + \xi \frac{d\eta}{d\xi} - (\xi^2+n^2)\eta=0 $$ and employ following transformation $$ \eta=\frac{y}{x^\alpha}, \quad \xi=\beta x^\gamma, $$ you should arrive at $$ \frac{d^2y}{dx^2} - \frac{(2\alpha-1)}{x}\frac{dy}{dx} - (\beta^2 \gamma^2 x^{2\gamma-2}+\frac{n^2\gamma^2-\alpha^2}{x^2})y=0. $$ To see the steps, I recommend to look into the book "Bowman: Introduction to Bessel functions, p.117", where it is done for standard Bessel equation.

Since $\eta$ satisfies the modified Bessel equation it follows that $\eta(\xi)=AK_n(\xi)+B I_n(\xi)$, $A,B$ being constants. Employing the transformations above, we have that the transformed Bessel equation has a general solution $$ y(x)=x^\alpha[A K_n(\beta x^\gamma)+B I_n(\beta x^\gamma)]. $$ Your specific equation is the case $\alpha=\frac12,\gamma=\frac12,\beta=2\sqrt{3},n=3$.

Note, that for non-integer $n$ we usually take the solution in a fom $$ y(x)=x^\alpha[A I_n(\beta x^\gamma)+B I_{-n}(\beta x^\gamma)]. $$


The following Wolfram code

y[x_] := Sqrt[x] K[2 Sqrt[3 x]];
u = (x^2 D[y[x], {x,2}] - (2 + 3 x) y[x]) 8 Sqrt[3]/t;
u/. x -> (t^2/12) // PowerExpand // Simplify

evaluates to

-(9 + t^2) K[t] + t (K'[t] + t K''[t])

which goes a long way to establishing the Bessel connection.

  • 1
    $\begingroup$ That does it! Thank you. I will write up my solution soon when I have more time. $\endgroup$ Dec 16, 2019 at 2:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.