I have a question regarding a PDE (Fokker-Planck) and change of variables. I have a problem deciding what route to take after I use the chain route. I have an expression $$\frac{\partial u}{\partial t} = \frac{\partial}{\partial x}\left(\frac{\partial}{\partial x}+x\right)u$$ and would like the variables substitution $$x = Xe^{-t} \text{ and } u=ve^t.$$ This give $$\frac{\partial v}{\partial t} = e^{2t} \frac{\partial^2 v}{\partial X^2}.$$ I thank you for your help.

  • $\begingroup$ What is the expanded form of $\dfrac{\partial}{\partial x}\left(\dfrac{\partial}{\partial x}+x\right)u$ ? $\endgroup$ – doraemonpaul Aug 2 '12 at 22:02
  • $\begingroup$ This question has been solved perfectly. Hope that the asker has been diving enough and accept the answer at an early date. $\endgroup$ – doraemonpaul Sep 10 '12 at 1:21

We have $$\partial_t u=\partial_x(\partial_x u+xu)=\partial_{xx}u+u+x\partial_xu.$$ We also also $v=e^{-t}u$, hence \begin{align} \partial_t v&=-e^{-t}u+e^{-t}\partial_t u\\ &=-e^{—t}u+e^{—t}\partial_{xx}u+e^{-t}u+e^{-t}x\partial_x u\\ &=e^{—t}\partial_{xx}u+e^{-t}x\partial_x u. \end{align} Now compute $\partial_X V$ with the chain rule. We have $v(X,t)=u(x,t),$ and $$\partial_X V=\partial_x u\frac{\partial x}{\partial X}+\partial_t u\frac{\partial t}{\partial X},$$ then substitute with the values of $\frac{\partial x}{\partial X}$ and $\frac{\partial t}{\partial X}$, then do the same for the second derivative.

| cite | improve this answer | |
  • $\begingroup$ >I thank you for your response $\endgroup$ – REULAND Aug 3 '12 at 17:42
  • $\begingroup$ > I cannot compute dV/dX and d^2V/dX^2 $\endgroup$ – REULAND Aug 3 '12 at 17:45
  • $\begingroup$ > I can use :differentiation of X=xu/V - find by X=xe^t and u=v*e^t $\endgroup$ – REULAND Aug 3 '12 at 17:53
  • $\begingroup$ > May I have a complete proof of dv/dt=e^(2*t) d^2v/dX^2 - Thanking you in advance for your help. $\endgroup$ – REULAND Aug 3 '12 at 17:58
  • $\begingroup$ I've added a step, but not all the details. $\endgroup$ – Davide Giraudo Aug 3 '12 at 18:10

We should take high precautions about changing variables of PDEs. Don't replay the mistakes like Why solving $\dfrac{\partial u}{\partial x}=\dfrac{\partial^2u}{\partial y^2}$ like this is wrong? !

Note that

$\dfrac{\partial u}{\partial t}=\dfrac{\partial}{\partial x}\left(\dfrac{\partial}{\partial x}+x\right)u$

$\dfrac{\partial u}{\partial t}=\dfrac{\partial}{\partial x}\left(\dfrac{\partial u}{\partial x}+xu\right)$

$\dfrac{\partial u}{\partial t}=\dfrac{\partial^2u}{\partial x^2}+x\dfrac{\partial u}{\partial x}+u$

Let $\begin{cases}x=Xe^{-t}\\T=t\end{cases}$ ,

Then $\begin{cases}X=xe^t\\T=t\end{cases}$

$\dfrac{\partial u}{\partial x}=\dfrac{\partial u}{\partial X}\dfrac{\partial X}{\partial x}+\dfrac{\partial u}{\partial T}\dfrac{\partial T}{\partial x}=e^t\dfrac{\partial u}{\partial X}$

$\dfrac{\partial^2u}{\partial x^2}=\dfrac{\partial u}{\partial x}\left(e^t\dfrac{\partial u}{\partial X}\right)=\dfrac{\partial u}{\partial X}\left(e^t\dfrac{\partial u}{\partial X}\right)\dfrac{\partial X}{\partial x}+\dfrac{\partial u}{\partial T}\left(e^t\dfrac{\partial u}{\partial X}\right)\dfrac{\partial T}{\partial x}=e^t\dfrac{\partial^2u}{\partial X^2}e^t=e^{2t}\dfrac{\partial^2u}{\partial X^2}$

$\dfrac{\partial u}{\partial t}=\dfrac{\partial u}{\partial X}\dfrac{\partial X}{\partial t}+\dfrac{\partial u}{\partial T}\dfrac{\partial T}{\partial t}=xe^t\dfrac{\partial u}{\partial X}+\dfrac{\partial u}{\partial T}$

$\therefore xe^t\dfrac{\partial u}{\partial X}+\dfrac{\partial u}{\partial T}=e^{2t}\dfrac{\partial^2u}{\partial X^2}+xe^t\dfrac{\partial u}{\partial X}+u$

$\dfrac{\partial u}{\partial t}=e^{2t}\dfrac{\partial^2u}{\partial X^2}+u$

Let $u=ve^t$ ,

Then $\dfrac{\partial u}{\partial X}=\dfrac{\partial(ve^t)}{\partial X}=e^t\dfrac{\partial v}{\partial X}$

$\dfrac{\partial^2u}{\partial X^2}=\dfrac{\partial}{\partial X}\left(e^t\dfrac{\partial v}{\partial X}\right)=e^t\dfrac{\partial^2v}{\partial X^2}$

$\dfrac{\partial u}{\partial t}=\dfrac{\partial(ve^t)}{\partial t}=e^t\dfrac{\partial v}{\partial t}+ve^t$

$\therefore e^t\dfrac{\partial v}{\partial t}+ve^t=e^{3t}\dfrac{\partial^2v}{\partial X^2}+ve^t$

$e^t\dfrac{\partial v}{\partial t}=e^{3t}\dfrac{\partial^2v}{\partial X^2}$

$\dfrac{\partial v}{\partial t}=e^{2t}\dfrac{\partial^2v}{\partial X^2}$

Of course the following PDE is then solved by using separation of variables:

Let $v=f(X)g(t)$ ,

Then $f(X)g'(t)=e^{2t}f''(X)g(t)$



$\begin{cases}g(t)=c_3(s)e^{-\frac{e^{2t}s^2}{2}}\\f(X)=\begin{cases}c_1(s)\sin Xs+c_2(s)\cos Xs&\text{when}~s\neq0\\c_1X+c_2&\text{when}~s=0\end{cases}\end{cases}$

$\therefore v=c_1X+c_2+\int_sc_3(s)e^{-\frac{e^{2t}s^2}{2}}\sin Xs~ds+\int_sc_4(s)e^{-\frac{e^{2t}s^2}{2}}\cos Xs~ds$ or $v=c_1X+c_2+\sum_sc_3(s)e^{-\frac{e^{2t}s^2}{2}}\sin Xs+\sum_sc_4(s)e^{-\frac{e^{2t}s^2}{2}}\cos Xs$

$ue^{-t}=c_1xe^t+c_2+\int_sc_3(s)e^{-\frac{e^{2t}s^2}{2}}\sin xe^ts~ds+\int_sc_4(s)e^{-\frac{e^{2t}s^2}{2}}\cos xe^ts~ds$ or $ue^{-t}=c_1xe^t+c_2+\sum_sc_3(s)e^{-\frac{e^{2t}s^2}{2}}\sin xe^ts+\sum_sc_4(s)e^{-\frac{e^{2t}s^2}{2}}\cos xe^ts$

$u=c_1xe^{2t}+c_2e^t+\int_sc_3(s)e^{t-\frac{e^{2t}s^2}{2}}\sin xe^ts~ds+\int_sc_4(s)e^{t-\frac{e^{2t}s^2}{2}}\cos xe^ts~ds$ or $u=c_1xe^{2t}+c_2e^t+\sum_sc_3(s)e^{t-\frac{e^{2t}s^2}{2}}\sin xe^ts+\sum_sc_4(s)e^{t-\frac{e^{2t}s^2}{2}}\cos xe^ts$

| cite | improve this answer | |
  • $\begingroup$ > Many thanks once again ! $\endgroup$ – REULAND Aug 8 '12 at 12:54
  • $\begingroup$ > du/dX instead of du/dx - then du/dX= d(ve^t)/dX = e^t dv/dX and d^2u/dX^2= d/dX(e^t dv/dX)= e^t d^2v/dX^2- du/dt=d(ve^t)/dt=e^tdv/dt +ve^t - How do you have : e^t dv/dt +ve^t= e^3*t d^2/dX^2 + v*e^t - I do not understand the way to obtain: e^3*t d^2v/dX^2 ? I thank you for your help ( I am a beginner in Maths ) $\endgroup$ – REULAND Aug 16 '12 at 13:08
  • $\begingroup$ @REULAND: Just substitute the results of $\dfrac{\partial u}{\partial X}$ , $\dfrac{\partial^2u}{\partial X^2}$ and $\dfrac{\partial u}{\partial t}$ into the previous PDE $\dfrac{\partial u}{\partial t}=e^{2t}\dfrac{\partial^2u}{\partial X^2}+u$ . Reminder that the term $\dfrac{\partial^2u}{\partial X^2}$ also contain the coefficient $e^{2t}$ , so we get $e^{3t}\dfrac{\partial^2v}{\partial X^2}$ . Please use TEX for your comment. $\endgroup$ – doraemonpaul Aug 16 '12 at 17:55

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.