I'm interested in the way of solving the following recurrence relation:

$$a_{n+1}=a_n'+a_1 a_n-b_1 b_n \\ b_{n+1}=b_n'+b_1 a_n+a_1 b_n$$

Where all $a_n,b_n$ are functions of $x$, and $'$ means derivative w.r.t. $x$. The initial functions $a_1$ and $b_1$ are known. Also, assume we have closed forms the derivatives: $a_1^{(n)}$ and $b_1^{(n)}$ in terms of $n$.

I expect there to be a closed form solution for this case, however, I have trouble deriving it.

One idea I had was to use operator notation and matrix notation, then we have:

$$\begin{bmatrix} a_{n+1} \\ b_{n+1} \end{bmatrix}= \begin{bmatrix} D+a_1 & -b_1 \\ D+b_1 & a_1 \end{bmatrix} \begin{bmatrix} a_n \\ b_n \end{bmatrix}$$

Which obviously can be written as:

$$\begin{bmatrix} a_{n+1} \\ b_{n+1} \end{bmatrix}= \begin{bmatrix} D+a_1 & -b_1 \\ D+b_1 & a_1 \end{bmatrix}^n \begin{bmatrix} a_1 \\ b_1 \end{bmatrix}$$

However, the $n$th power of a matrix is found through its eigenvalues, and I'm not sure how to find eigenvalues for the operator matrix.

We can also rewrite this in a more clear way:

$$\begin{bmatrix} a_{n+1} \\ b_{n+1} \end{bmatrix}= \left(\begin{bmatrix} 1 & 0 \\ 1 & 0 \end{bmatrix} \frac{d}{dx} +\begin{bmatrix} a_1 & -b_1 \\ b_1 & a_1 \end{bmatrix} \right)^n \begin{bmatrix} a_1 \\ b_1 \end{bmatrix}$$

And use binomial sum? Not sure. How to correctly write binomials sum with a differential operator inside?

Or maybe there's a more simple way to obtain the closed form?

I would prefer to use strictly real methods if possible, thought I kind of doubt it could work without complex numbers.


By letting $z_n = a_n + i b_n$ the given equations can be coupled as

$$z_{n+1} = z_n' + z_1 z_n = (D+z_1)z_n = (D+z_1)^n z_1$$

$$z_2=z_1'+z_1^2$$ $$z_3=z_1''+3 z_1 z_1'+z_1^3$$ $$z_4=z_1'''+3 z_1'^2 + 4 z_1 z_1''+6 z_1' z_1^2 + z_1^4$$ and the general form for $z_n$ is given by Faà di Bruno's formula.

  • 1
    $\begingroup$ I should have mentioned that I don't want to use the complex numbers. But considering the matrix I have got, it's all just complex numbers in disguise, so I don't believe any real methods are possible here $\endgroup$ – Yuriy S Nov 7 '18 at 7:57

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.