4

I'm currently reading a book in Multivariable Calculus, and there is a section on Applications of Calculus to Physics - Poisson's Equation. It states the following:

We have the 3D version: Let $$u(x) = \iiint_{\Bbb{R}^3}\frac{p(x+y)}{|y|}d^3y$$, where $p$ is a $C^2$ function on $\mathbb{R}^3$, that vanished outside a bounded set. Then $u$ is of class $C^2$ and $\nabla^2u = -4\pi p$.

The 2D analog: Let $u(x) = \int p(x+y)\log{|y|}d^2y$, where $p$ is a $C^2$ function on $\mathbb{R}^2$, that vanished outside a bounded set. Then $u$ is of class $C^2$ and $\nabla^2u = 2\pi p$.

I was wondering how we could generalize this to 4 dimensions - or even n dimensions (if its possible).

My conjecture for the 4D version is: Let $$u(x) = \iiiint_{\Bbb{R}^4}\frac{p(x+y)}{|y|^2}d^4y$$, where $p$ is a $C^2$ function on $\mathbb{R}^4$, that vanished outside a bounded set. Then $u$ is of class $C^2$ and $\nabla^2u = -4\pi^2 p$.

Is this conjecture correct? If not, what would the correct analog be, and if so, how could we prove it?

K.defaoite
  • 13,890
kirky49
  • 63

2 Answers2

3

In THIS ANSWER, I developed the Green (or Green's if you prefer) function, $G(\vec x|\vec y)$ for the $n$-dimensional Inhomogeneous Helmholtz equation,

$$\nabla^2 G_k(\vec x|\vec y)+k^2 G_k(\vec x|\vec y)=-\delta (\vec x-\vec y)\tag1$$

where $\delta(\vec x)$ is the Dirac Delta. Solution to $(1)$ can be written

$$G_k(\vec x|\vec y)=\frac i4 \left(\frac{k}{2\pi |\vec x-\vec y|}\right)^{n/2-1}H_{n/2-1}^{(1)}(k|\vec x-\vec y|)\tag2$$

where $H^{(1)}_{\alpha}(z)$ is the Hankel function of the first kind and order $\alpha$.

Using the same approach for the case $k=0$, we find the Green function for Poisson's equation, $\nabla^2 G_0(\vec x|\vec y)=-\delta(\vec x-\vec y)$ is given by

$$G_0(\vec x|\vec y)=\frac{\Gamma(n/2-1)}{4\pi^{n/2}|\vec x-\vec y|^{n-2}}\tag3$$



ASIDE DISCUSSION:

We could have arrived at $(2)$ by taking the limit as $k\to 0$ of $(2)$. To do so, we use the small argument (i.e., small $k|\vec r-\vec r'|$) approximation for the Hankel function of the first kind and order $n/2-1$, $n>2$, which is given by

$$H_{n/2-1}^{(1)}(k|\vec r-\vec r'|) = -i\frac{\Gamma(n/2-1)}{\pi} \left(\frac2{k|\vec r-\vec r'|}\right)^{n/2-1}+O(k^{n/2-1})\tag{A1}$$

Using $(A1)$ in $(2)$, we find that the small argument approximation for $G_k(\vec x|\vec y)$ for $n>2$ is given by

$$G_k(\vec x|\vec y)=\frac{\Gamma(n/2-1)}{4\pi^{n/2}|\vec x-\vec y|^{n-2}}+O(k^{n-2})\tag {A2}$$

If we let $k\to 0$ in $(A2)$, we see that solution to the $n$-dimensional Poisson Equation $\Delta G_0(\vec x|\vec y)=-\delta(\vec x-\vec y)$ is

$$G_0(\vec x|\vec y)=\frac{\Gamma(n/2-1)}{4\pi^{n/2}|\vec x-\vec y|^{n-2}}$$



Using $(3)$, we find the solution to Poisson's equation $\nabla^2 u(\vec x)=p(\vec x)$ can be written as

$$\begin{align} u(\vec x)&=\int_{\mathbb{R}^n}p(\vec y)G(\vec x|\vec y)\,d^n\vec y\\\\ &=\int_{\mathbb{R}^n}p(\vec y)\frac{\Gamma(n/2-1)}{4\pi^{n/2}|\vec x-\vec y|^{n-2}} \,d^n\vec y\\\\ &=\int_{\mathbb{R}^n}p(\vec x+\vec y)\frac{\Gamma(n/2-1)}{4\pi^{n/2}|\vec y|^{n-2}} \,d^n\vec y \end{align}$$



EXAMPLES:

Example $1$ ($n=3$):

For $n=3$, $\Gamma(n/2-1)=\pi^{1/2}$ and $G_0(\vec x|\vec y)=\frac{1}{4\pi |\vec x-\vec y|}$ is the familiar Green function and we have

$$\begin{align} u(\vec x)&=\int_{\mathbb{R}^3}\frac{p(\vec y)}{4\pi |\vec x-\vec y|}\,d^3\vec y\\\\ &=\int_{\mathbb{R}^3}\frac{p(\vec x+\vec y)}{4\pi |\vec y|}\,d^3\vec y \end{align}$$


Example $2$ ($n=4$):

For $n=4$, $\Gamma(n/2-1)=1$ and $G_0(\vec x|\vec y)=\frac{1}{4\pi^2 |\vec x-\vec y|^2}$ we have

$$\begin{align} u(\vec x)&=\int_{\mathbb{R}^4}\frac{p(\vec y)}{4\pi^2 |\vec x-\vec y|^2}\,d^4\vec y\\\\ &=\int_{\mathbb{R}^4}\frac{p(\vec x+\vec y)}{4\pi^2 |\vec y|^2}\,d^4\vec y \end{align}$$


Example $3$ ($n=5$):

For $n=5$, $\Gamma(n/2-1)=\sqrt\pi/2$ and $G_0(\vec x|\vec y)=\frac{1}{4\pi^2 |\vec x-\vec y|^2}$ we have

$$\begin{align} u(\vec x)&=\int_{\mathbb{R}^4}\frac{p(\vec y)}{8\pi^2 |\vec x-\vec y|^3}\,d^5\vec y\\\\ &=\int_{\mathbb{R}^4}\frac{p(\vec x+\vec y)}{8\pi^2 |\vec y|^3}\,d^5\vec y \end{align}$$

Mark Viola
  • 184,670
  • Sorry, I've never seen the Helmholtz equation, so I'm not sure. I'm also not familiar with some of the notation in your answer. I'm just wondering whether the 4D analog I conjectured is correct, with respect to the 2D and 3D versions I stated, and how to prove it. – kirky49 Apr 15 '21 at 23:11
  • Hi. Thanks for your help. I'm not really familiar with what $G(r|r')$ means, and I'm not sure where the $\delta$ and $\Gamma$ are coming from. If possible, could you please use the terminology that I used in the question to correct/prove my conjecture? i.e define $u(x)$ and solve for $\nabla^2u$ in terms of $p$, like how my book did it for the 2D and 3D version. I really appreciate it! – kirky49 Apr 16 '21 at 00:32
  • I'm afraid I don't understand your answer from the link, as I've never studied that function before, and I'm not sure how to transfer your answer from that into the same terminology as I used. I'm also unsure what you mean by integrate $G$ with $4\pi p$ to get $u$. – kirky49 Apr 16 '21 at 01:26
  • Hi again. I've made extensive modifications to the solution herein to (1) use your notation for spatial variables $\vec x$ and $\vec y$. I've also added several links as references that support the treatise (e.g., Dirac Delta, Gamma function, Hankel function, Green function). And I've added examples for $n=3$, $n=4$, and $n=5$ dimensions. What else can I do to give you the best answer I can? – Mark Viola Apr 16 '21 at 14:47
1

The $n$-dimensional version (for $n \ge 2$) is that $$ u(x)=\int_{\mathbb{R}^n} p(x+y) \, R(|y|) \, dy $$ satisfies $\nabla^2 u = p$ if the function $R(r)$ is such that $1/R'(r)$ equals the $(n-1)$-dimensional surface area of the sphere with radius $r$ in $\mathbb{R}^n$. That is, $$ R'(r) = \frac{1}{A_n r^{n-1}} , $$ where $A_n = 2 \pi^{n/2} \, / \, \Gamma(n/2)$ is the area of the unit sphere in $\mathbb{R}^n$ (see this question, for example).

(I won't prove this here, but the phrase to search for if you want to learn more is the fundamental solution for the Laplace operator.)

For the exceptional case $n=2$, this gives $R'(r)=\frac{1}{2\pi r}$ so that $R(r)=\frac{1}{2\pi} \ln r$ (plus a constant, which is conventionally taken to be zero).

For $n \ge 3$, we get instead $$ R(r) = \frac{-1}{(n-2) A_n r^{n-2}} $$ (plus a constant, which is conventionally taken to be zero). So your conjecture for $n=4$ is correct, since we indeed have $(n-2)A_n = 2 A_4 = 2 \cdot 2 \pi^2 = 4 \pi^2$.

Hans Lundmark
  • 55,038
  • How does this result differ from the one in the solution that I posted? – Mark Viola Apr 16 '21 at 13:33
  • @MarkViola: The result is the same, of course! But since the OP seemed to have some trouble understanding your answer, I thought I could try another explanation. It's perhaps a bit overkill to go via the general solution for the Helmholtz equation with Hankel functions, when you can rather easily prove the formula for the Poisson equation directly (although I was too lazy to include the proof here). Anyway, feel free to downvote if you don't like it. :-) – Hans Lundmark Apr 16 '21 at 14:03
  • 1
    Hans, I rarely down vote and would not down vote your solid post. And you're correct. It is just as easy to find the Green function for Poisson's Equation directly. I'll add that to my post. – Mark Viola Apr 16 '21 at 14:08
  • Could you please reiterate why my Conjecture is correct? I can't seem to connect my conjecture with your answer – kirky49 Apr 21 '21 at 01:38
  • @kirky49: According to the formula in my answer, the choice $R(r)=-1/(4\pi r^2)$ makes $$v(x) = \iiiint_{\Bbb{R}^4} p(x+y) , R(|y|) , d^4y = - \frac{1}{4 \pi^2} \iiiint_{\Bbb{R}^4} \frac{p(x+y)}{|y|^2} , d^4y$$ satisfy $\nabla^2 v = p$, so your function $u$, which is $u = -4 \pi^2 v$, satisfies $\nabla^2 u = -4 \pi^2 p$ as you conjectured. – Hans Lundmark Apr 21 '21 at 05:28