# Differential equation - homogenous solution

How can I verify that the differential equation:

$$y''_n +(2n\coth x)y'_n+(n^2 -1) y_n =0$$

has homogeneous solution

$$y_n = \left(\frac{1}{\sinh x} \frac{d}{dx}\right)^n (Ae^x + Be^{^-x})$$

for degree $n \in N$

How to start this problem? Should I calulate first and second derivative from d/dn * y_n and later try to insert into equation? I don´t really understand how it can be verified. I will be grateful for all help.

• You should be able to simplify that expression for $y_n$ a little. The sinh part will just be to the nth power and then the exponentials have a simple form under repeated differentiation. From there plugging in the expression should work. Commented Feb 10, 2018 at 21:35

In order to prove the result for arbitrary values of $n$ we can use an induction argument. Assume that the hypothesis holds for a nominal value $n$ and then derive steps to prove that it also holds for $n+1$. This requires that we first verify the hypothesis for low values.

Proving for $n=0$

Suppose that $n=0$. The stated homogenous solution is, $$y_0 = Ae^x + Be^{-x} \hspace{1cm} \tag{1}$$

and the stated differential equation is, $$y_0'' - y_n = 0 \hspace{1cm} \tag{2}$$

Taking the first and second derivatives of (1) yields, \begin{align} y_0' =& Ae^{x} - Be^{-x} \\ y_0'' =& Ae^{x} + Be^{-x} \end{align} and so, $$y_0'' - y_0 = 0$$

This is that same as stated in (2), and so we can assert that the solution is true for $n=0$.

Proving for $n=1$

Suppose now that $n=1$. The differential equation is, $$y_1'' + (2\coth{x})y_1' = 0$$

the proposed solution is, \begin{align} y_1 =& \frac{1}{\sinh{x}} \frac{d}{dx}(Ae^{x} + Be^{-x}) \\ =& \frac{1}{\sinh{x}} (Ae^{x} - Be^{-x}) \hspace{1cm} \tag{3} \end{align}

Differentiating (3) gives, \begin{align} y_1' =& (B-A)\mbox{csch}^2{x} \\ y_1'' =& 2(A - B) \coth(x)\mbox{csch}^2(x) \end{align}

and so,

\begin{align} y_1'' + (2\coth{(x)})y_1' =& 2(A - B) \coth(x)\mbox{csch}^2(x) + (2\coth{(x)})(B-A)\mbox{csch}^2{(x)} \\ =&2(A - B) \coth(x)\mbox{csch}^2(x) - 2(B-A)\coth{(x)}\mbox{csch}^2{(x)} \\ =& 0 \end{align}

as required.

Asserting the hypothesis for arbitrary $n$

Suppose that the hypothesis holds for $n$. We then have that, $$y_n'' + (2n \coth)y_n' + (n^2 - 1)y_n = 0 \tag{4}$$ and also a solution, $$y_n = \left(\mbox{csch} \frac{d}{dx}\right)^n (A e^{x} + Be^{-x}) \tag{5}$$

We are free to differentiate Equation (4) with respect to x. This will come in useful futher down. $$y_n''' + (2n \coth(x)) y_n' - (2n\mbox{csch}^2(x)) y_n' + (n^2 - 1)y_n' = 0 \tag{6}$$

The hypothesis that we are tying to prove is that, $$y_{n+1}'' + (2(n+1)\coth(x))y_{n+1}' + ((n+1)^2 - 1)y_{n+1} = 0 \tag{7}$$

In order to reach the hypothesis we will first compute the values of $y_{n+1}, y_{n+1}'$ and $y_{n+1}''$, using the solution in Equation (5). \begin{align} y_{n+1} =& \left(\mbox{csch}(x) \frac{d}{dx} \right) y_n = \mbox{csch}(x)y_n' \tag{8}\\ y_{n+1}' =& \left( \frac{d}{dx} \mbox{csch}(x) \right) y_n' + \mbox{csch}(x) y_n'' \tag{9}\\ y_{n+1}'' =& \left( \frac{d^2}{dx^2} \mbox{csch}(x) \right) y_n' + 2 \left( \frac{d}{dx} \mbox{csch}(x) \right) y_n'' + \mbox{csch}(x) y_n'''\tag{10} \end{align}

Notice that equation (10) has a triple derivative in, which needs to be dealt with. Rearranging equation (6) gives that, $$y_n''' = (2n\mbox{csch}^2(x)) y_n' - (2n \coth(x)) y_n' - (n^2 - 1)y_n' \tag{11}$$

If we now put Equation (11) into Equation (10), and then combine that with Equations (9) and (8) to put into the left hand side of Equation (7); some rearranging gives, $$\left((2n+1)\mbox{csch}^3(x) - (1 + 2n) \mbox{csch}(x)\coth^2(x) - (n^2 -1)\mbox{csch}(x) + n(n+2) \mbox{csch}(x) \right) y_n' = 0 \tag{12}$$

Which proves the result.

Note: I used Wolfram Alpha to verify Equation (12).

• I assumed that $n=1, A=1, B =1$ But I am not sure if I can make assumption for A,B. I calculated derviatives without this assumption and the results of the derivatives were complex. After this assumption my differential equation has form $$y''_n +(2\coth x)y'_n =0$$ and $$y_n = \left(\frac{1}{\sinh x} \frac{d}{dx}\right) (e^x + e^{^-x})$$ Then $y_n$ has a form of $y_n =-2cothx^2$. Then I calculated $y_n´$, $y_n´´$ and I inserted it into differential equation. The result which I obtained was $-4csch(x)^4$ which is not equal to 0 (DE doesnt have homogenous solution). Is it correct logic?
– Vid
Commented Feb 9, 2018 at 14:35
• You only need to choose $n$. $A$ and $B$ are constant and since differentiation is a linear operation they will become coefficients. It looks to me as though the equations are incorrect. Commented Feb 9, 2018 at 18:17
• Can you tell me where I went wrong? Commented Feb 10, 2018 at 21:11
• @Daniel Beale the coefficient of y is $(n^2-1)$ which is $-1$ for n=0. This means that eqn 2 should say $y_0''=y_0$. Commented Feb 10, 2018 at 21:25
• I´m sorry @Daniel Beale, earlier my comment was partly saved. After inserting $n=0$ my differential equation has form: $$y_n´´-y_n=0 (1)$$. (- instead of +). and $$y_0´=Ae^x-Be^{-x}$$ $$y_0´´=Ae^x+Be^{-x}$$ After inserting it into (1) equation i got result $$0=0$$ What are the further steps which have to be done? Or this is the final prove that this differential equation has the homogenous solution.
– Vid
Commented Feb 10, 2018 at 21:39