0
$\begingroup$

Well, I never had to show something like this which is why I'm having quite a hard time to get this one done. I basically know what I have to do but I am not capable of solving it properly.

Given for example: $U = 0.25, F = 0.01, \Delta t = 0.1$

$ \begin{align*} u_k &= U + \underbrace{u_{k-1}(1-U)}_\text{A} \,\underbrace{e^{-\Delta t /F}}_\text{B} \end{align*} $

Well, since $U$ is constant, the only interesting things are $A$ and $B$.

I know that $A$ itself converges to $0$. The problem is I don't know how I can show that analytically. $B$ also would regulate the whole term that is added to $U$ so in the end it will determine the value where $u_k$ will converge to.

As far as I understood it $u_0 = U$.

Could someone explain to me a technique that would allow me to solve such an equation?

So basically, what is $\lim_{k\rightarrow\infty} u_k$.

$\endgroup$
1
$\begingroup$

General Formula for an Arithmetico-Geometric Sequence

Since $U$, $\Delta t$ and $F$ are constant, your recursion reduces to an arithmetico-geometric sequence: $$u_k=a\cdot u_{k-1}+b$$ If we let $r=\dfrac{b}{1-a}$, we get: $$u_n=a^n(u_o-r)+r$$


Specific Example

In your particular case, we have: $a=\dfrac{0.75}{e^{10}}$ and $b=u_0=0.25$, such that: $$u_n=\left(\dfrac{0.75}{e^{10}}\right)^n\left(0.25-\dfrac{0.25}{1-\frac{0.75}{e^{10}}}\right)+\dfrac{0.25}{1-\frac{0.75}{e^{10}}}$$ Since $0<\dfrac{0.75}{e^{10}}<1$, we have: $$\lim_{n\to\infty}u_n=0+\dfrac{0.25}{1-\frac{0.75}{e^{10}}}\simeq 0.250009$$


Proof of the General Formula

Consider the sequence: $$u_k=a\cdot u_{k-1}+b$$ and let $w_{k-1}=u_k-u_{k-1}$, such that: $$w_k=u_{k+1}-u_k=(a\cdot u_k+b)-(a\cdot u_{k-1}+b)=a(u_k-u_{k-1})=a\cdot w_{k-1}$$ Thus $w_n$ is a geometric sequence of ratio $a$ and first term $w_0=u_1-u_0=(a-1)u_0+b$. From the definition of $w_k$ as the difference of $u_k$ and $u_{k-1}$, we have that: $$\sum_{k=0}^{k=n}w_k=u_n-u_0$$ And from the definition of $w_k$ as a geometric sequence: $$\sum_{k=0}^{k=n}w_k=w_0\dfrac{a^n-1}{a-1}=\dfrac{(a-1)u_0+b}{a-1}(a^n-1)$$ Let $r=\dfrac{b}{1-a}$, and equate the two sums above: $$u_n-u_0=(u_0-r)(a^n-1)$$ And finally we recover the general formula: $$u_n=a^n(u_0-r)+r$$

$\endgroup$
  • $\begingroup$ Okay that means I can do that with anything as long I can rewrite it to a form like $u_k = a \cdot u_{k-1} + 1$, right? Would you call something like this something like "linear" recursive equation? And is a more general approach too far-reaching? I mean like if there is an approach I could use to solve even more complex recursive equations? I don't expect you to answer this my question is answered with what you provided already but I wouldn't mind some more explanation. :) Thank you in any case! $\endgroup$ – displayname May 4 '15 at 20:41
  • $\begingroup$ As long as you can rewrite it as $u_k=a\cdot u_{k-1}+b$, with $a$ and $b$ constant, yes. This is not the most general approach, but when you're given a recursion and need to find if it converges, it's always a good thing to check if it is not by any chance a "nice" recursion (arithmetic, geometric or arithmetico-geometric sequences). Because in these cases, you can easily obtain a closed-form formula, and the limit follows naturally. $\endgroup$ – Demosthene May 4 '15 at 20:46
  • $\begingroup$ Okay, thank you. I guess you just "need an eye" for things like that on the long run. If things like that come up more often I think I'll have to learn a little more about such equations. $\endgroup$ – displayname May 4 '15 at 20:53

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.