I'm currently studying the Bonnor solution starting with this paper on Black Diholes. The metric is given by :

$$ ds^2 = \left(1-\frac{2Mr}\Sigma\right)^2 \left[-dt^2 + \frac{\Sigma^4}{(\Delta + (M^2 +a^2)\sin^2 \theta)^3}\left(\frac{dr^2}{\Delta} + d\theta^2\right)\right] + \frac{\Delta \sin^2\theta}{\left(1-\frac{2Mr}\Sigma\right)^2}d\phi^2 \tag{1}$$

where $\Delta = r^2 + 2 M r - a^2 = (r-r_+)(r-r_-)$ and $\Sigma = r^2 - a^2 \cos^2\theta$.

I'm getting the "picture" behind the conical defect on a manifold: take a piece of paper, cut off a "wedge" and then glue the two sides together -- a cone is obtained (see the figure for clarity).

enter image description here

In more concrete terms, any circumference of center O in the original manifold doesn't close anymore (it spans an angle $ \theta < 2\pi$ because of the cut), while identifying the two sides it can still close. The presence of a deficit angle is hence signaled if the ratio between the circumference and the radius is not $2\pi$.

What I would do to compute the deficit angle in the context of pseudo-Riemannian manifolds: take some cylindrical-like coordinates with z coordinate coinciding with the axis of the "cone", then the radius is the proper distance of a radial path starting from the axis ($r=0$) to $r=R$; as a circumference one uses the proper length of a curve around the given axis (i.e. in the identified direction), taken $r=R$. Take the ratio of such quantities, then send $R$ to zero.

But is finding such an exact change of coordinates always an efficient strategy? Since eq.(1) is given in Boyer-Lindquist-like coordinates, I guess there is a more general approach to the problem, which doesn't require any particular choice of coordinates.

Luckily the paper offers two examples, each for a different axis of simmetry, from which we can learn :

$$ \delta_0 = 2\pi - \left|\frac{\Delta \phi \space d \sqrt{ g_{\phi \phi}}}{\sqrt{ g_{\theta\theta}}\ d\theta}\right| \tag{4}$$

$$ \delta_{r_+} = 2\pi - \left|\frac{\Delta \phi \space d \sqrt{ g_{\phi \phi}}}{\sqrt{ g_{rr}}\ dr}\right| \tag{5}$$

Where it is said that the $\phi$ coordinate is identified with period $\Delta \phi$.

Apparently both numerators look like what I was thinking about. But I can't understand the factors in the denominators. Moreover I don't get whether the "d" in both formulae is a true derivative or not.

Can anyone provide an insight on these formulae?

UPDATE: I tried moving to cylindrical coordinates. Since eq.(1), as said, is expressed in Boyer-Lindquist coordinates, the change of coordinates involves of course just the ($r$,$\theta$) coordinates:

\begin{cases} \rho = \sin{\theta} \sqrt{r^2 - 2 M r + a^2} \\ z = (r-M) \cos{ \theta} \end{cases}

Let's compute the deficit as explained, in these coordinates: I'm taking an infinitesimal circumference $\gamma_{\epsilon}$ around the z-axis ($\theta = 0$). The proper length of such thing would be (remember that $ \phi \in (0, \Delta \phi) $)

$$ \int_{\gamma_{\epsilon}} \sqrt{g_{\mu \nu} dx^\mu dx^\nu} = \int_0^{\Delta \phi} \sqrt{g_{\phi \phi}}\ d\phi \vert_{\rho = \epsilon} $$

and the proper radius

$$ \int_0^\epsilon \sqrt{g_{\rho \rho}}\ d\rho$$

Hence the defect of the angle would be something like :

$$ \delta_0 = 2\pi - \lim_{\epsilon \to 0} \frac{\int_0^{\Delta \phi} \sqrt{g_{\phi \phi}}\ d\phi \vert_{\rho = \epsilon}}{\int_0^\epsilon \sqrt{g_{\rho \rho}}\ d\rho} $$

Now for sloppiness we drop the integration sign, substituting the limit with the infinitesimal version of the integrand. So I'm interpreting the "d"s in the paper as just a compact way to express this limit.

However my result looks more like eq.(5) rather than eq.(4). Any ideas of what's going on?


1 Answer 1


I figured out what happened: I was trying to "see" the cone as a 3D structure, "emerging" in the $\theta = 0$ direction, hence forgetting it is actually a 2D plane, which has undergone an angular identification; in this way I couldn't even generalize the procedure for the conical singularity in the "axis" $r = r_+$, since it was lacking an intuitive geometrical interpretation.

I think the following procedure could still be possible in Weyl/Cylindrical coordinates, however this would imply a mixing of both $(r,\theta)$ coordinates, introducing unncessary complications in the limits. Using Boyer-Lindquist coordinates is actually a simple way to get things done.

Since we're looking for a 2D manifold, let's take the induced metric for $t, r = const.$ :

$$ ds^2 = \left(1-\frac{2Mr}\Sigma\right)^2 \left[\frac{\Sigma^4}{(\Delta + (M^2 +a^2)\sin^2 \theta)^3} d\theta^2\right] + \frac{\Delta \sin^2\theta}{\left(1-\frac{2Mr}\Sigma\right)^2}d\phi^2 $$

and since one seeks $g_{\phi \phi} = 0$, from eq.(1) one has to let $\theta \to 0 \lor \pi$. In both limits, one sees that the metric simply becomes :

$$ ds^2 = C \left( d\theta^2 + \sin^2 \theta \space d\phi^2 \right)$$

where C is a constant. This has the more familiar face of a 2D-Euclidean manifold with polar-like coordinates (just use the "$\theta$" as a radius). One can finally take the proper length of circumference as said before :

$$ L = \int_0^{\Delta \phi} \sqrt{g_{\phi \phi}}\ d\phi \vert_{\theta = \epsilon} = C \Delta \phi \sin \epsilon $$

and the radius :

$$ R = \int_0^{\epsilon} \sqrt{g_{\theta \theta}} \space d \theta = C \epsilon $$

Now taking the limit, one recovers the result of the paper:

$$ \lim_{\epsilon \to 0} \frac{L}{R} = \lim_{\epsilon \to 0} \frac{C \Delta \phi \sin\epsilon}{C \epsilon} = \Delta \phi$$

In analogue fashion, one can look at the other piece of the axis of symmetry, i.e. $r = r_+$. Let's consider the induced metric for $t, \theta = const.$ and take the first leading order of the metric in the limit $r \to r_+$. What you get is:

$$ ds^2 = C \left( \frac{f(r) dr^2}{\left (a^2 + M^2 \right)^3} + \frac{d\phi^2}{f(r)} \right) $$

where C is a constant (that depends on $\theta$, which is fixed) and $f(r) = \frac{a^4}{2 \sqrt{a^2 + M^2} \left(r-r_+ \right)}$.

Now it's simply a matter of redifining R to make the metric look like this:

$$ ds^2 = C \left( \frac{dR^2}{\left (a^2 + M^2 \right)^3} + \frac{R^2 d\phi^2}{4A^2} \right) $$

Since the 2D metric in polar coordinates is recovered just by finding $R = R(r)$, the OP's eq.(5) is still valid - by taking integrals in the place of differentials as mentioned in the "UPDATE" section.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.