3
$\begingroup$

Does this P.D.E: $$\nabla\cdot\left( \frac{ \nabla u}{u} \right)+a\, \Delta u+b\,u=0 \hspace{3cm} (*)$$ have a variational structure? Here $a$ and $b$ are constants.

In other words, the question I am asking is:

Does there exist a functional such that the corresponding Euler–Lagrange equation is (*)?

$\endgroup$
3
$\begingroup$

I)

Answer: Yes, choose functional $$\tag{1} S[u]~:=~\int_{\Omega\subseteq \mathbb{R}^n} d^nx~{\cal L}(u,\nabla u),\qquad d^nx~:=~dx_1 \wedge \ldots \wedge dx_n,$$
where $$\tag{2} {\cal L}(u,\nabla u)~:=~\frac{1}{2}\left(a+\frac{1}{u}\right)^2(\nabla u)^2 -\frac{ab}{2}u^2 -bu. $$

Sketched proof: The functional derivative is

$$ \frac{\delta S[u]}{\delta u} ~=~\frac{\partial {\cal L}(u,\nabla u)}{\partial u} -\nabla \cdot \frac{\partial {\cal L}(u,\nabla u)}{\partial (\nabla u)}$$ $$\tag{3}~=~-\left(a+\frac{1}{u}\right) \left[-\frac{1}{u^2}(\nabla u)^2 +\left(a+\frac{1}{u}\right)\nabla^2u +bu\right],$$

which leads to the Euler-Lagrange equation

$$\tag{4}\nabla\cdot\left(\frac{1}{u}\nabla u\right) +a\nabla^2u +bu~=~0\quad \vee\quad u~=~-\frac{1}{a}.$$

II)

More generally, let us for fun consider the normalized equation $$\tag{5} \nabla^2u+ f(u)(\nabla u)^2+g(u)~=~0, $$ where $f$ and $g$ are given functions of $u$. The question is now whether we can find a variational principle for eq. (5)? The answer is again Yes.

Let $F$ and ${\cal V}$ be antiderivatives (aka. primitive or indefinite integrals) of $f$ and $ge^{2F}$, respectively, i.e.

$$\tag{6}F^{\prime}(u)~=~f(u) \quad\text{and}\quad {\cal V}^{\prime}(u)~=~g(u)e^{2F(u)}.$$

We now multiply the normalized equation (5) with an integrating factor

$$\tag{7}\lambda(u)~:=~ e^{2F(u)}. $$

Define the Lagrangian density $$\tag{8}{\cal L}(u,\nabla u) ~:=~\frac{1}{2}e^{2F(u)}(\nabla u)^2-{\cal V}(u).$$ It is straightforward to show that the Euler-Lagrange equation for the Lagrangian density (8) is eq. (5).

Sketched proof:

$$ \frac{\delta S[u]}{\delta u} ~=~\frac{\partial {\cal L}(u,\nabla u)}{\partial u} -\nabla \cdot \frac{\partial {\cal L}(u,\nabla u)}{\partial (\nabla u)}$$ $$\tag{9}~=~-e^{2F(u)}\left[ \nabla^2u+ f(u)(\nabla u)^2+g(u) \right].$$

$\endgroup$
  • $\begingroup$ Thank you for the comment. Before checking this functional, I'm curious how you found it. Also, what does $d^n x$ mean? The n-times differentiation w.r.t $x$ of $x\in\mathbb{R}^n$? $\endgroup$ – LCH Dec 21 '14 at 5:16
  • $\begingroup$ I updated the answer. $\endgroup$ – Qmechanic Dec 21 '14 at 17:29
  • $\begingroup$ It makes me surprised to know that this PDE has a variational structure. Thank you so much! $\endgroup$ – LCH Dec 21 '14 at 20:05

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.