5
$\begingroup$

I want to know whether or not $$ \limsup_{\epsilon \rightarrow 0^+} \int_{D}\dfrac{1}{\sqrt{(x-(1-\epsilon))^2 +y^2}} \frac{1}{\sqrt{1-\sqrt{x^2+y^2}}} \, dx \, dy <+\infty.$$

Here $D $ denotes the disk $\{(x,y)\in\mathbb{R}^2:x^2+y^2<1\}$.

Mathematica indicates that the limit is finite

In:= NIntegrate[(1/Sqrt[(x - 0.99999999999)^2 + 
         y^2])*(1/Sqrt[1 - Sqrt[x^2 + y^2]]), {x, -1, 1}, {y, -Sqrt[1 - x^2], Sqrt[1 - x^2]}]


Out:= 13.3727774796526

but along with the warning:

Numerical integration converging too slowly; suspect one of the 
following: singularity, value of the integration is 0, highly 
oscillatory integrand.

The $L^1$-norm of $\dfrac{1}{\sqrt{(x-(1-\epsilon))^2 +y^2}}$ is uniformly bounded, and $\dfrac{1}{\sqrt{1-\sqrt{x^2+y^2}}}$ is integrable. However I don't know how to handle the singularity when the two are combined.

$\endgroup$
3
  • $\begingroup$ Post deleted until I figure out whether Patrick Li's objection is fatal. $\endgroup$ Commented Aug 15 at 23:55
  • $\begingroup$ I'm sorry, I screwed up and dropped a 2, and my answer is unfixable. I still think the integral is convergent, and I might try and leave another answer, but I'm going to leave this alone for a few days and hope someone else will answer it in the meantime. $\endgroup$ Commented Aug 16 at 0:50
  • 2
    $\begingroup$ I think integrability at $\epsilon =0$ (not exactly what the OP is asking, of course) can be established by changing variables to $(x,y)=(s+1,y)$ and doing a Taylor expansion of the denominator of the second factor about $(s,y)=(0,0)$ (obviously the only point that poses problems). When we then integrate in polar coordinates (wrt $(s,y)$), we should be dealing with essentially $\int d\varphi\int dr\,/\sqrt{r\cos\varphi +(r^2/2)\sin\varphi}$, and now the $r$ integral is of order $\log \cos\varphi$ near $s=0$, which is integrable in $\varphi$. $\endgroup$ Commented Aug 16 at 23:26

1 Answer 1

6
$\begingroup$

Transforming to polar coordinates, $$ \int_{D}\dfrac{1}{\sqrt{(x-(1-\epsilon))^2 +y^2}} \frac{1}{\sqrt{1-\sqrt{x^2+y^2}}} \, dx \, dy = $$ $$ \int_{0}^{1} dr \frac{1}{\sqrt{1-r} } \int_{0}^{2\pi } d\phi \frac{1}{\sqrt{1-2 ((1-\epsilon)/r) \cos \phi + (1-\epsilon)^2 /r^2 } } $$ The inner, $\phi $, integral can be evaluated in closed form, $$ I = \int_{0}^{2\pi } d\phi \frac{1}{\sqrt{1-2 ((1-\epsilon)/r) \cos \phi + (1-\epsilon)^2 /r^2 } } =\frac{4}{1+ ((1-\epsilon)/r)} K\left( \frac{4(1-\epsilon)/r }{(1+((1-\epsilon)/r))^2 }\right) $$ where $K$ is the complete elliptic integral of the first kind. $K$ diverges logarithmically when its argument approaches unity, i.e., when $r\rightarrow 1-\epsilon $, such that $$ I= -2 \log |(1-\epsilon )-r| \ \ + \ \ \mbox{terms regular at} \ \ r=1-\epsilon $$ This singularity is, of course, integrable. Moreover, when $\epsilon \rightarrow 0$, the $r$-integrand has the singularity $\log (1-r)/\sqrt{1-r} $, which remains integrable.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .