How to compute this distribution?

156 Views Asked by At

My question refers to this answer.

I was hoping someone could explain in more detail the following reasoning.

It remains to observe that $\Delta v$ is the distribution composed of

  • the linear measure on $\partial Q$
  • $-\sqrt{2}$ times the linear measure on the diagonals of $Q$

This follows from considering the discontinuities of the normal derivative of $v$ across the aforementioned lines; elsewhere $v$ is harmonic. One can also save the trouble of calculating the factor of $-\sqrt{2}$ by using the fact that $\int_{\mathbb R^2}\Delta v=0$.

I do not see how these measures arise from the discontinuities of the derivative. I do see, however, why it is zero elsewhere. (Because it is harmonic there, so the Laplacian vanishes.)

I am familiar with the theory of distributions as outlined in Folland's book, so I don't need any formation about what distributions are. I just want to know how to compute this specific one.

1

There are 1 best solutions below

6
On BEST ANSWER

Consider the cube given $Q=[-1,1]^2$ and the diagonals $D_1,D_2$. These divide the cube into four triangles, call them $A_i$. Now we have for any $u\in C_c^\infty(\mathbb{R}^2)$ that

$$ \left\langle \Delta v, u \right\rangle = \int_{\mathbb{R}^2} v\Delta u dx = -\int_{Q} \nabla v\cdot \nabla u dx $$ where the first equallity is by definition and the second is an integration by parts (made possible since $v$ is Lipschitz and is supported in $Q$). Now we rewrite this last integral as

$$ \int_Q \nabla v\cdot\nabla u dx = \sum_{i=1}^4 \int_{A_i} \nabla v \cdot \nabla u dx \tag{1} $$ We can apply again the divergence theorem (or Green's formula) to this since $v$ is smooth in each $A_i$, but now we have to take into account the boundary terms: In other words we get

$$ \int_{A_i} \nabla v\cdot \nabla u dx = -\int_{A_i} (\Delta v)u + \int_{\partial A_i} \frac{\partial v}{\partial \nu_i} u \sigma =\int_{\partial A_i} \frac{\partial v}{\partial \nu_i} u d \sigma ,\tag{2} $$ where $\nu_i$ is the outward unit normal to $\partial A_i$ and $\sigma$ is the surface (or in this case length) measure.

Order the $A_i$ so that $\nabla v=(1,0), (0,1), (-1,0), (0,-1)$ in $A_1,A_2,A_3,A_4$ respectively (in other words $A_1$ is the triangle to the left and you enumerate counterclockwise). Now notice that if $T$ represents a rotation of $\pi/2$ then

$$ T( \nabla v|_{A_i}) = \nabla v |_{A_{i+1}} \quad \text{and} \quad T(\nu_i)=\nu_{i+1} $$ With these we get that $\frac{\partial v}{\partial \nu_i} = \frac{\partial v}{\partial \nu_j}$ for every $i,j$. Take then $i=1$ and we compute that this derivative is $-1$ in $\partial A_1 \cap \partial Q$ and $1/\sqrt{2}$ in $\partial A_1 \cap Q$. With this we get

$$ \int_{\partial A_i} \frac{\partial v}{\partial \nu_i} u d \sigma = -\int_{\partial A_i \cap Q} u d \sigma + \frac{1}{\sqrt{2}}\int_{\partial A_i\cap Q} u d \sigma. $$ Adding these we get, noting that each of the two line segements in $\partial A_i\cap Q$ is shared with exactly one other $\partial A_j$ we get

$$ \sum_{i=1}^4 \int_{\partial A_i \cap Q} \frac{\partial v}{\partial \nu_i} u d \sigma = \sqrt{2}\sum_{i=1}^2\int_{D_i} u d \sigma. $$

In a similar fashion

$$ \sum_{i=1}^4\int_{\partial A_i \cap \partial Q} \frac{\partial v}{\partial \nu_i} u d \sigma = -\int_{\partial Q} u d \sigma. $$

This, together with (1) and (2), give

$$ \left\langle \Delta v, u \right\rangle = \int_Q u d \sigma - \sqrt{2} \int_{D_i} u d \sigma $$

which is what you want.