Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

i want to solve the followning second order non linear PDE :

$$\frac{\partial V}{\partial t}(t,x) + g(x)\frac{\partial V}{\partial x}(t,x) + q(x)\frac{\partial^{2} V}{\partial x^{2}}(t,x) + h(x) =0 $$

with boundary condition $V(T,x) = \Phi(x)$ where $t\le T$.

I heard that the method of characteristics can work in order to determine $V(t,x)$. However, i have never met this method before. Could someone give me a quick quide ?


share|cite|improve this question
Yes, someone could: Method of characteristics. – Did Jan 28 '12 at 8:26
@Didier: Dear Didier, That article only really describes the application of the method to first order equations, though. Regards, – Matt E Jan 28 '12 at 12:25
Dear peter, You describe your PDE as non-linear, but looking at it, it seems linear to me. What am I missing? Regards, – Matt E Jan 28 '12 at 12:27
Firstly, thanks for the immediate responce. @Matt E: Dear Matt it is a non linear PDE, probably my fault that i didnt write the functions g,q,h in a better way. Dear Didier, i have seen this article too, but i am looking for second orfer PDEs. In fact i have to solve a system of two PDEs of this forms, but i suppose that if i manage to solve one of them i would probably solve the system too. I was thinking to give $V(t,x)$ the form of a Cole-Hopf factorization , i.e. $V(t,x)=e^{\alpha f(t)}$ and then get a Burgers-like PDE. However, i coudn't find simething instructive in order to begin my wor – peter Jan 28 '12 at 12:41
Your PDE is linear. In fact, linearity is meant w.r.t. the unknown $V$, not w.r.t. the variables $(x,t)$: this means that a differential operator $\mathcal{L}$ is said to be linear iff $\mathcal{L}[\alpha V_1+\beta V_2]=\alpha \mathcal{L}[V_1] + \beta \mathcal{L}[V_2]$, and this is the case for your operator $\mathcal{L}=q \frac{\partial^2}{\partial x^2} + g\frac{\partial}{\partial x} + \frac{\partial}{\partial t}$, where $q,g,h$ depend on $(x,t)$. Neverthless, your PDE is inhomogeneous because of the term $h$. Hence, you have a second order, inhomogeneous linear PDE of parabolic type. – Pacciu Mar 28 '12 at 16:11

Case $1$: $g(x)=q(x)=0$

Then $\dfrac{\partial V(t,x)}{\partial t}+h(x)=0$

$\dfrac{\partial V(t,x)}{\partial t}=-h(x)$


$V(T,x)=\Phi(x)$ :



$\therefore V(t,x)=\Phi(x)+h(x)T-h(x)t=\Phi(x)+h(x)(T-t)$

Case $2$: $g(x)\neq0$ and $q(x)=0$

Then $\dfrac{\partial V(t,x)}{\partial t}+g(x)\dfrac{\partial V(t,x)}{\partial x}+h(x)=0$

$\dfrac{\partial V(t,x)}{\partial t}+g(x)\dfrac{\partial V(t,x)}{\partial x}=-h(x)$

This belongs to a PDE of the form

The general solution is $V(t,x)=C\biggl(\int^x\dfrac{1}{g(x)}dx-t\biggr)-\int^x\dfrac{h(x)}{g(x)}dx$

$V(T,x)=\Phi(x)$ :




$\therefore V(t,x)=\Phi\biggl(\biggl(\int^x\dfrac{1}{g(x)}dx-T\biggr)^{-1}\biggl(\int^x\dfrac{1}{g(x)}dx-t\biggr)\biggr)+\int^{\left(\int^x\frac{1}{g(x)}dx-T\right)^{-1}\left(\int^x\frac{1}{g(x)}dx-t\right)}\dfrac{h(x)}{g(x)}dx-\int^x\dfrac{h(x)}{g(x)}dx$

Case $3$: $q(x)\neq0$

Then $\dfrac{\partial V(t,x)}{\partial t}+g(x)\dfrac{\partial V(t,x)}{\partial x}+q(x)\dfrac{\partial^2V(t,x)}{\partial x^2}+h(x)=0$

It is possible that the subsititution $V(t,x)=V_c(t,x)+V_p(t,x)$ can make the inhomogeneous linear PDE becomes homogeneous linear PDE if $V_p(t,x)$ can be found.

For this question, the form of $V_p(t,x)$ is not difficult to guess, just the particular solution of the ODE $g(x)\dfrac{dV(x)}{dx}+q(x)\dfrac{d^2V(x)}{dx^2}+h(x)=0$ , i.e. $V_p(t,x)=-\int^xe^{-\int^x\frac{g(x)}{q(x)}dx}\int^x\dfrac{h(x)}{q(x)}e^{\int^x\frac{g(x)}{q(x)}dx}~dx~dx$

So let $V(t,x)=V_c(t,x)-\int^xe^{-\int^x\frac{g(x)}{q(x)}dx}\int^x\dfrac{h(x)}{q(x)}e^{\int^x\frac{g(x)}{q(x)}dx}~dx~dx$ ,

Then $\dfrac{\partial V_c(t,x)}{\partial t}+g(x)\dfrac{\partial V_c(t,x)}{\partial x}+q(x)\dfrac{\partial^2V_c(t,x)}{\partial x^2}=0$

Of course we use separation of variables:

Let $V_c(t,x)=T(t)X(x)$ ,

Then $T'(t)X(x)+g(x)T(t)X'(x)+q(x)T(t)X''(x)=0$




$\therefore V(t,x)=\sum_sC_1(s)e^{tf_s(s)}X_{s,1}(x,s)+\sum_sC_2(s)e^{tf_s(s)}X_{s,2}(x,s)+\int_sC_3(s)e^{tf_g(s)}X_{g,1}(x,s)~ds+\int_sC_4(s)e^{tf_g(s)}X_{g,2}(x,s)~ds-\int^xe^{-\int^x\frac{g(x)}{q(x)}dx}\int^x\dfrac{h(x)}{q(x)}e^{\int^x\frac{g(x)}{q(x)}dx}~dx~dx$

or $V(t,x)=\sum_sC_1(s)e^{tf_s(s)}X_{s,1}(x,s)+\sum_sC_2(s)e^{tf_s(s)}X_{s,2}(x,s)+\sum_sC_3(s)e^{tf_g(s)}X_{g,1}(x,s)+\sum_sC_4(s)e^{tf_g(s)}X_{g,2}(x,s)-\int^xe^{-\int^x\frac{g(x)}{q(x)}dx}\int^x\dfrac{h(x)}{q(x)}e^{\int^x\frac{g(x)}{q(x)}dx}~dx~dx$

Then subsititute $V(T,x)=\Phi(x)$ into the above general solutions to eliminate part of the arbitrary parts to get the final answer.

The suffix $s$ stand for special cases while suffix $g$ stand for general cases. The special cases means that the values of $s$ that lead $q(x)X''(x)+g(x)X'(x)+f(s)X(x)=0$ will have the forms of their general solution differ from the current form of the general solution of most values of $s$ in $q(x)X''(x)+g(x)X'(x)+f(s)X(x)=0$ .

Note that different $q(x)$ , $g(x)$ and $f(s)$ may result the special cases appearing in the different values of $s$ . Like the previous question Indication on how to solve the heat equations with nonconstant coefficients $(q(x)=-x^2 , g(x)=-2x)$ , choose $f(s)=-\dfrac{4\pi^2s^2+1}{4}$ will result the only special case appearing in $s=0$ . We should choose the form of $f(s)$ wisely so that the special cases appear in the nice values of $s$ .

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.