Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Think of $t$ and $r$ as two independent variables.

  • Suppose $E$ be a function of $r$ and $V~$ be a function of $(t,r)$ such that both go to $0$ at $r=0$.

  • There exists a positive function $M(r)$ such that $M(0)=0$ and $V(t,r) = -\dfrac{M(r)}{R(t,r)}$ where $R$ is another positive function such that $R(0,r)=r$.

  • Let $p(r) = \dfrac{E(r)}{V(0,r)}$ be a function regular at $r=0$ such that $p(0) \in (-\infty,1)$.

  • Also define a function $a$ of $r$ such that, $a(r) = \dfrac{M(r)}{\dfrac{4}{3}\pi r^3}$. Then $a$ is also a positive definite function with a well-defined value at $r=0$.

  • Define $\alpha = a(0)$

Now look at this differential equation,

$$\frac{\dot{R}^2}{2} + V(t,r) = E(r)$$

Apparently this differential equation has a solution of the form,

$$\frac{t}{t_0} = \sqrt{\frac{\alpha}{a(r)}}\frac{F(p(r))}{F(p(0))} \left [1 - \left ( \dfrac{R(t,r)}{r} \right)^{\dfrac{3}{2}}~\cdot~\dfrac{F\left(~~ \dfrac{p(r)R(t,r)}{r} \right) }{F(p(r))} \right ] $$

where $t_0 = \sqrt {\dfrac{3}{8\pi \alpha}} F(p(0))$

and the function $F$ is defined over the interval $(-\infty,1)$ as,

$$F(x) = \left\{ \begin{array}{c c} -\frac{\sqrt{1-x}}{x} - \frac{1}{(-x)^{\frac{3}{2}}} \tanh^{-1} \left [ \sqrt{\frac{x}{x-1}} \right ] & x<0 \\ \frac{2}{3} & x =0 \\ \frac{1}{x^{\frac{3}{2}}}tan^{-1} \left [ \sqrt{\frac{x}{1-x}} \right ] - \frac{\sqrt{1-x}}{x} & 0<x<1 \end {array} \right. $$

How does one get the above solution?

share|cite|improve this question
What do you mean by a "positive definite function"? And how does that reconcile with the notion that $M(0) = 0$? – Willie Wong Oct 11 '10 at 22:45
@Willie Thanks for pointing out the typo. I have corrected it. Any help with solving this differential equation? – Anirbit Oct 13 '10 at 6:26
I've fixed your latex: the problem was that you needed to \-escape a \{ and all \\, and that the < sign has to be written in some cases using an HTML escape (for silly reasons! Of course, in comments the rules are different...) – Mariano Suárez-Alvarez Oct 13 '10 at 20:26
What is $\dot{R}$, is it $\partial R/\partial t$? Also, there seems to be a lot of interdependence among the definitions; can you tell us what is given for a particular instance of the problem, and what is to be determined? Better yet, if you could provide the original source or motivation for this problem, it might make things clearer. – Rahul Oct 13 '10 at 20:45
@Mariano Thanks for correcting the LaTeXing. @Rahul Yes $\dot{R}$ is partial derivative of $R$ with respect to it. I didn't understand the second part of your query. I think I have completely defined all the quantities in question. You see any ambiguities? This is taken from a paper and I am giving you the reference if that helps, "Strength of naked singularities in Tolman-Bondi spacetimes" by R.P.A.C Newman in Class. Quantum Grav.3 (1986) 527-539 The third page of the paper has this. Will be happy to get back any help. – Anirbit Oct 14 '10 at 4:33
up vote 4 down vote accepted

I'm sort of just guessing here (partly based on the solution already found). By explicitly plugging in $V(t,r) = - \frac{M(r)}{R(t,r)}$, you arrive at the ordinary differential equation (for each fixed $r$) for $R$ as

$$ \dot{R}^2 - \frac{2M}{R} = 2 E $$

where $M$ and $E$ are constants in time. Now, re-scale the original equation by $t = \lambda s$ and $R = \mu \rho$. Then $\partial_t R = \frac{\mu}{\lambda} \partial_s \rho$. Then you can solve $(\frac{\mu}{\lambda})^2 = \frac{2M}{\mu} = 2E$ to reduce the equation to

$$ \dot{\rho}^2 - \frac{1}{\rho} = 1 $$

(the weights $\mu$ and $\lambda$ will, roughly speaking, give you the weights $p(r)$ and $a(r)$ in your question). Now note that this scaling degenerates if $E = 0$. In the case that $E = 0$, the equation can be solved by quadrature:

$$ \dot{x}^2 = x^{-1} \Rightarrow \sqrt{x} dx = dt \Rightarrow x^{3/2} \sim t $$

This gives the solution to the homogeneous case. In the inhomogeneous case, you take that as a sort of integrating factor: assume that $\rho^{3/2} f(\rho) \sim t$, this implies that

$$ \dot{\rho} \left( \rho^{3/2} f(\rho) \right)' = 1 $$

we plug this into the equation, which we first re-arrange as

$$ \dot{\rho} = \sqrt{\frac{1}{\frac{\rho}{1+\rho}}} $$

and we conclude that

$$ \sqrt{\frac{\rho}{1+\rho}} = \frac{d}{d\rho}( \rho^{3/2} f(\rho)) $$

and you solve this by directly integrating it. I think this $f$ you find should be exactly the $F$ you wrote down above (I didn't check it myself). Note that due to the singular weight $\rho$ which degenerates as $\rho \to 0$, you will have to separately integrate in the regime where $\rho > 0$ and $\rho < 0$. The existence and uniqueness of solution is guaranteed by the theory of Fuchsian ODEs.

share|cite|improve this answer
Thanks a lot for this help. Are you defining $f$ as, $f(\rho) = \frac{kt}{\rho ^ \frac{3}{2}}$ ? (for some constant $k$) Since there are no other replies I am giving away the bounty to you. (Though I am yet to understand your solution completely) – Anirbit Oct 19 '10 at 6:18
Actually I think I defined it with $k = 1$. The constant $k$ can be absorbed into the (implicit) definition of $\rho$, by rescaling $\rho$; so it is not too important. The reason I wrote $\sim t$ instead of $=t$ is just because there can be a constant term: $f(\rho) = (kt + C) / \rho^{3/2}$. who doesn't change the derivative. (In fact, I think in the final answer, to match the boundary conditions there may need to be a $C$.) – Willie Wong Oct 19 '10 at 10:06

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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