I have a nonlinear Diffeq:

$$\frac{d^2x}{dt^2}+\beta \frac{dx}{dt}+\epsilon \times e^{- \lambda x} = f(t) $$

where $f(t)$ is a function that is known, and $\beta$ and $\lambda$ are constants that are known. Also, we know that $\epsilon$ is a constant parameter that is small.

I first need to obtain the zero order solution $x_0$, before finding the first order solution $x_1$

The first thing that I need to do is to use asymptotic expansions to obtain solutions of order $\epsilon=0$ and (TYPO)

Note that general solution for f(t) that will have two unknown constants.

UPDATE: After the first order term is solve, it needs to be plugged back in. The exponential needs to be linearized and things should start cancelling out. I am not sure how to do this, I just know this is what needs to be done.

UPDATE2: Correction, $\epsilon = 1$ was a typo. It should be $\epsilon^1$ I need to find a solution in the form:

$$x(t)=x_0(t)+\epsilon^1x_1(t)+\epsilon^2x_2 (t) + ... $$

So initially, $\epsilon$ needs to be set to 0 in order to obtain $x_0$. To find $x_1$, I need $\epsilon^1$

UPDATE3: I know now that I need to plug:

$$x=x_0+\epsilon_1x_1 $$ back into the original equation


$$\frac{d^2}{dt^2}(x_0+\epsilon_1x_1) + \beta\frac{d}{dt}(x_0+\epsilon_1x_1)+\epsilon \times exp(-\lambda(x_0+\epsilon_1x_1)) $$


$$\frac{d^2}{dt^2}x_0+\frac{d^2}{dt^2}\epsilon_1x_1+\beta \frac{d}{dt}x_0 +\beta \frac{d}{dt}\epsilon_1 x_1+\epsilon \times exp(-\lambda x_0))+\epsilon \times exp(-\lambda \epsilon_1 x_1)$$

I think then the $x_0$ terms may cancel with f(t) or something like that? It may be some sort of approximation.

I still need to linearize the exponential. Any help is appreciated.

Update4: Taking the solution a little but further...

We know that:

$$\frac{d^2x_0}{dt^2}+\beta \frac{dx_0}{dt} = f(t) $$

So, those terms all cancel. And now we have: $$\frac{d^2}{dt^2}\epsilon_1x_1 +\beta \frac{d}{dt}\epsilon_1 x_1+\epsilon \times exp(-\lambda(x_0+\epsilon_1x_1))=0$$

But we dont want $\epsilon^2$ terms, to part of the exponential goes away as well.

We are left with: $$\frac{d^2}{dt^2}\epsilon_1x_1 +\beta \frac{d}{dt}\epsilon_1x_1+\epsilon \times exp(-\lambda x_0)=0$$

Where we know $x_0$. This now means that the exponential is no longer a function of arbitrary x. I feel like the solution should be trivial now, but I am having a hard time finding it. Any ideas?

Can this be solved with the method of undetermined coefficients?

UPDATE5: Well I have updated this problem several times with very little response. As a latch ditch effort, is there anyone who can offer any advice on how to solve:

$$\frac{d^2x_1}{dt^2} +\beta \frac{dx_1}{dt} + exp{-\lambda x_0}=0$$

where $x_0$ is known

  • $\begingroup$ Have you done a similar problem before? A first order equation? Something with a simpler nonlinear term, e.g. $\epsilon \cdot x^2$ instead of $\epsilon \cdot e^{-\lambda x}$? $\endgroup$ – Hans Engler Dec 7 '14 at 1:15
  • $\begingroup$ Look up your notes or a textbook and (re)do a simpler problem. Btw the solution to this problem will still involve an integral involving $f(t)$. $\endgroup$ – Hans Engler Dec 7 '14 at 1:18
  • $\begingroup$ "If you have a problem that you cannot solve, then there is a simpler problem that you cans solve. Find such a problem and solve it." That's advice from George Polya ("How to solve it"). en.wikipedia.org/wiki/How_to_Solve_It $\endgroup$ – Hans Engler Dec 7 '14 at 1:26
  • $\begingroup$ Google3 for "regular perturbation ode". There are several good intro articles that will come up. The ones from ucl.ac.uk, cims.nyu.edu and utwente.nl are all good places to start. Good luck. $\endgroup$ – Hans Engler Dec 7 '14 at 1:39
  • $\begingroup$ You don't have any specifications on $ f(t) $ ? $\endgroup$ – David Cardozo Dec 8 '14 at 19:03

A possible solution is as follows:

Write the equation as

${\frac {d^{2}}{d{t}^{2}}}x \left( t \right) +\beta\,{\frac {d}{dt}}x \left( t \right) +\epsilon\,{{\rm e}^{-\lambda\,x \left( t \right) }} =f_{{0}} \left( t \right) +f_{{1}} \left( t \right) \epsilon $

Now look for a solution of the form

$x \left( t \right) =x_{{0}} \left( t \right) +x_{{1}} \left( t \right) \epsilon $

Replacing this solution in the equation we have

${\frac {d^{2}}{d{t}^{2}}}x_{{0}} \left( t \right) + \left( {\frac {d^{ 2}}{d{t}^{2}}}x_{{1}} \left( t \right) \right) \epsilon+\beta\, \left( {\frac {d}{dt}}x_{{0}} \left( t \right) + \left( {\frac {d}{dt }}x_{{1}} \left( t \right) \right) \epsilon \right) +\epsilon\,{ {\rm e}^{-\lambda\, \left( x_{{0}} \left( t \right) +x_{{1}} \left( t \right) \epsilon \right) }}=f_{{0}} \left( t \right) +f_{{1}} \left( t \right) \epsilon $

expanding the exponential we obtain

${\frac {d^{2}}{d{t}^{2}}}x_{{0}} \left( t \right) +\beta\,{\frac {d}{d t}}x_{{0}} \left( t \right) + \left( {\frac {d^{2}}{d{t}^{2}}}x_{{1}} \left( t \right) +\beta\,{\frac {d}{dt}}x_{{1}} \left( t \right) +{ {\rm e}^{-\lambda\,x_{{0}} \left( t \right) }} \right) \epsilon=f_{{0} } \left( t \right) +f_{{1}} \left( t \right) \epsilon $

Then we deduce that

${\frac {d^{2}}{d{t}^{2}}}x_{{0}} \left( t \right) +\beta\,{\frac {d}{d t}}x_{{0}} \left( t \right) =f_{{0}} \left( t \right) $


${\frac {d^{2}}{d{t}^{2}}}x_{{1}} \left( t \right) +\beta\,{\frac {d}{d t}}x_{{1}} \left( t \right) +{{\rm e}^{-\lambda\,x_{{0}} \left( t \right) }}=f_{{1}} \left( t \right) $

The solution at zero order is

$x_{{0}} \left( t \right) =\int _{0}^{t}\! \left( \int _{0}^{\tau}\!f_{ {0}} \left( \sigma \right) {{\rm e}^{\beta\,\sigma}}{d\sigma}+C_{{1}} \right) {{\rm e}^{-\beta\,\tau}}{d\tau}+C_{{2}} $

Replacing this solution in the equation at first order we obtain a linear equation which can be solved formally as

$x_{{1}} \left( t \right) =\int _{0}^{t}\! \left( \int _{0}^{\tau}\!(f_{ {1}} \left( \sigma \right) - {{\rm e}^{-\lambda\,x_{{0}} \left( \sigma \right) }} ){{\rm e}^{\beta\,\sigma}}{d\sigma}+C_{{3}} \right) {{\rm e}^{-\beta\,\tau}}{d\tau}+C_{{4}}$

  • $\begingroup$ See update3 please $\endgroup$ – Jackson Hart Dec 10 '14 at 21:14
  • $\begingroup$ See update 4 too please $\endgroup$ – Jackson Hart Dec 10 '14 at 22:58
  • $\begingroup$ Can the last equation I have me solved with undetermined coeffs? $\endgroup$ – Jackson Hart Dec 11 '14 at 1:59

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.