Linearized equation of PDE

41 Views Asked by At

I am finding some trouble in calculating the Linearized equation for even Laplacian. The $p$th Laplacian is defined as $-\Delta_pv= \text{div}(|Dv|^{p−2}Dv)$.

I know the linearized $p$ th Laplacian operator is $L_v(\phi)= -\text{div}(|Dv|^{p−2}|D\phi| + (p − 2)|Dv|^{p−4}|(Dv \cdot D\phi)Dv)$. I know the definition of it but I am unable to calculate $I'$ explicitly where $I$ is associated functional. It will be extremely helpful if one can give an idea how to calculate it explicitly or give some reference where such kind of calculations has been done.

1

There are 1 best solutions below

2
On BEST ANSWER

First of all your definition of the $p$-Laplacian is not correct. The $p$-Laplacian is $$\tag{$\ast$}\Delta_pu(x) = \operatorname{div} (\vert D u \vert^{p-2} D u )$$ not $\operatorname{div} (\vert D u \vert^{p-2} \vert D u \vert )$ (the second doesn't make sense). You have a similar issue with the linearisation of the $p$-Laplacian. Also, you write $\Delta_p u=-\operatorname{div} (\vert D u \vert^{p-2} D u )$ instead of $(\ast)$ which is a non-standard sign convention---this is not a huge problem, but I will use the more standard definition $(\ast)$ in my answer.

To compute the linearisation of $\Delta_p$ around $v$ in the direction $\phi$ (denoted $L_v\phi$), you need to compute the power series expansion of $\Delta_p(v+\varepsilon \phi)$ as $\varepsilon\to0^+$. Then, by definition, $$\Delta_p(v+\varepsilon \phi) = \Delta_pv+ \varepsilon L_v\phi + O(\varepsilon^2) $$ using Big-O notation.

Now, we have that \begin{align*}\Delta_p (v + \varepsilon \phi) &= \operatorname{div} \big (\vert D v+\varepsilon D\phi \vert^{p-2} ( D v +\varepsilon D\phi )\big )\\ &= \operatorname{div} (\vert D v+\varepsilon D\phi \vert^{p-2} D v )+\varepsilon\operatorname{div} (\vert D v+\varepsilon D\phi \vert^{p-2} D\phi ) \end{align*} Next, \begin{align*} \vert D v+\varepsilon D\phi \vert^2&=\vert Dv\vert^2+2\varepsilon Dv\cdot D\phi + \varepsilon^2\vert D\phi\vert^2. \end{align*} Now, using that $(1+x)^\alpha=1+\alpha x +O(x^2)$ as $x\to 0$, we have that \begin{align*} \vert D v+\varepsilon D\phi \vert^{p-2} &= \big ( \vert Dv\vert^2+2\varepsilon Dv\cdot D\phi + \varepsilon^2\vert D\phi\vert^2\big )^{\frac{p-2}2}\\ &=\vert Dv\vert^{p-2} \bigg (1+2\varepsilon \frac{Dv\cdot D\phi}{\vert D v \vert^2} + O(\varepsilon^2) \bigg)^{\frac{p-2} 2 }\\ &=\vert Dv\vert^{p-2} \bigg (1+(p-2)\varepsilon \frac{Dv\cdot D\phi}{\vert D v \vert^2} + O(\varepsilon^2) \bigg)\\ &= \vert Dv\vert^{p-2} +(p-2)\varepsilon \vert Dv\vert^{p-4}Dv\cdot D\phi + O(\varepsilon^2). \end{align*} Hence, \begin{align*} &\hspace{-2em}\Delta_p (v + \varepsilon \phi) \\ &= \operatorname{div} (\vert D v\vert^{p-2} D v )+(p-2)\varepsilon\operatorname{div} (\vert Dv\vert^{p-4}(Dv\cdot D\phi )D v )+\varepsilon\operatorname{div} (\vert D v \vert^{p-2} D\phi ) +O(\varepsilon^2)\\ &=\Delta_pv + \varepsilon \operatorname{div}\big (\vert D v \vert^{p-2} D\phi +(p-2) \vert Dv\vert^{p-4}(Dv\cdot D\phi )D v \big ) +O(\varepsilon^2), \end{align*} which gives $$L_v\phi =\operatorname{div}\big (\vert D v \vert^{p-2} D\phi +(p-2) \vert Dv\vert^{p-4}(Dv\cdot D\phi )D v \big ). $$