Convolution with vectors

698 Views Asked by At

I have what I suspect is a simple problem. Yet, I find myself stuck.

My equation describes the electric field at a point $\mathbf{r}$ due to cumulative polarization $\mathbf{P}(\mathbf{r^\prime})$ at points $\mathbf{r^\prime}$, and looks like this:

$$ \mathbf{E}(\mathbf{r}) = \int \left\{\frac{3(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] - \frac{\mathbf{P}(\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^3}\right\}d\mathbf{r^\prime} $$

I want to eliminate the $\mathbf{r^\prime}$ by expressing $\mathbf{E}(\mathbf{r})$ as a convolution of two functions of the vector $\mathbf{r}$.

Wikipedia tells me that the convolution is defined as: $$ (f * g )(t) = \int_{-\infty}^\infty f(\tau) g(t - \tau) \, d\tau $$ So it seems there should be a way to express my equation as a convolution with respect to $\mathbf{r}-\mathbf{r^\prime}$. However, I'm not sure if this applies to vectors, or how to deal with the dot product in my equation above.

Any clues would be really appreciated!

Hugh

1

There are 1 best solutions below

4
On

We begin with splitting the integral into two terms: $$ \mathbf{E}(\mathbf{r}) = \int \left\{\frac{3(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] - \frac{\mathbf{P}(\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^3}\right\}d\mathbf{r^\prime} \\ = \int \frac{3(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] d\mathbf{r^\prime} - \int \frac{\mathbf{P}(\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^3} d\mathbf{r^\prime} $$

The second integral can easily be written as a convolution: $$ \int \frac{\mathbf{P}(\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^3} d\mathbf{r^\prime} = \mathbf{P}(\mathbf{r}) * \frac{1}{|\mathbf{r}|^3} $$

The first integral is more difficult. The $i$th coordinate is somewhat easier: $$ \left( \int \frac{3(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] d\mathbf{r^\prime} \right)_i = \int \frac{3(x_i-x_i^\prime)}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] d\mathbf{r^\prime} \\ = \int \mathbf{P}(\mathbf{r^\prime})\cdot\frac{3(x_i-x_i^\prime)(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5} d\mathbf{r^\prime} = \mathbf{P}(\mathbf{r}) \stackrel{\cdot}{*} \frac{3 x_i \mathbf{r}}{|\mathbf{r}|^5} $$ where the operator $\stackrel{\cdot}{*}$ denotes that both a convolution and a dot product are made: $$ \mathbf{P}(\mathbf{r}) \stackrel{\cdot}{*} \frac{3 x_i \mathbf{r}}{|\mathbf{r}|^5} = \sum_j P_j(\mathbf{r}) * \frac{3 x_i x_j}{|\mathbf{r}|^5} $$

Thus, $$ \int \frac{3(\mathbf{r}-\mathbf{r^\prime})}{|\mathbf{r}-\mathbf{r^\prime}|^5}[\mathbf{P}(\mathbf{r^\prime})\cdot(\mathbf{r}-\mathbf{r^\prime})] d\mathbf{r^\prime} = \sum_i \left( \mathbf{P}(\mathbf{r}) \stackrel{\cdot}{*} \frac{3 x_i \mathbf{r}}{|\mathbf{r}|^5} \right) \mathbf{e}_i = \sum_i \sum_j P_j(\mathbf{r}) * \frac{3 x_i x_j}{|\mathbf{r}|^5} \mathbf{e}_i $$ so $$ \mathbf{E}(\mathbf{r}) = \sum_i \sum_j P_j(\mathbf{r}) * \frac{3 x_i x_j}{|\mathbf{r}|^5} \mathbf{e}_i + \mathbf{P}(\mathbf{r}) * \frac{1}{|\mathbf{r}|^3} $$


Isn't the integrand the expansion of something? I think that I recognize it. The non-expanded expression might be easier to write as a pure convolution.