14

A post on reddit inspired me and another contributor to wonder if there is a metric that produces a $1/r$ force law, or logarithmic potential, at least in the large-$r$ limit (if not exactly). Is any such metric known? Or is there a reason why it can't exist?

I would ordinarily look to 2+1D spacetime, because the Newtonian gravitational force in 2D is $1/r$, but that has been calculated and there is no gravitational force whatsoever in 2+1D GR. The spacetime around a point mass is flat with an angle defect.

Qmechanic
  • 201,751
David Z
  • 76,371
  • 1
    Comment to the post (v2): Are you requiring that the metric satisfies Einstein's vacuum field equations in the bulk? If not it becomes an exercise in determining a matter distribution/stress-energy-momentum tensor $T_{\mu\nu}$ that gives rise to a $1/r$ force law in the bulk. – Qmechanic Sep 01 '16 at 14:27
  • @Qmechanic Hm, well, I suppose the intent was to have a metric satisfying the vacuum field equations outside some region of finite extent in at least one dimension. Though since I didn't mention that, I probably shouldn't add it as a requirement at this late stage. – David Z Sep 01 '16 at 14:56
  • Gravity in $2+1$ dimension with a cosmological constant is pretty much this. It is easy to replicate in $3+1$ dimension by considering the metric of a cosmic string with a cosmological constant, I think. – Slereah Jan 16 '18 at 08:14
  • Concerning vacuum solutions, see e.g. https://physics.stackexchange.com/q/200020/2451 , https://physics.stackexchange.com/q/680025/2451 and links therein. – Qmechanic Mar 29 '23 at 08:12

1 Answers1

7

I'm guessing you want your metric to be spherically symmetry and to tend asymptotically to flat spacetime. In that case you want something like:

$$ \mathrm ds^2 = -a(r)\mathrm dt^2 + \frac{\mathrm dr^2}{b(r)} + \mathrm d\Omega^2 $$

where both $a(r)$ and $b(r)$ have to tend to one for large $r$.

A $1/r$ force law is going to require that the Christoffel symbol $\Gamma^r_{tt}$ is approximately $1/r$. One quick thrash of Mathematica later and I get:

$$ \Gamma^r_{tt} = -\tfrac{1}{2}b(r)~\frac{\mathrm da(r)}{\mathrm dr} $$

As a quick check, for the Schwarzschild metric we expect $\Gamma^r_{tt}$ is approximately $1/r^2$ to give the inverse square law. For this metric:

$$ a(r) = b(r) = 1-\frac{2GM}{c^2r} $$

So:

$$ \Gamma^r_{tt} = -\left(1-\frac{2GM}{c^2r}\right)\frac{GM}{c^2r^2} $$

and in the limit of $r \rightarrow\infty$ we get $\Gamma^r_{tt}\propto 1/r^2$ as we expect. So far so good.

So you just need to find two functions $a(r)$ and $b(r)$ such that both tend to unity at large $r$ and:

$$ b(r)~\frac{\mathrm da(r)}{\mathrm dr} \approx \frac{1}{r}$$

for large $r$. Typically you'd look for functions like $1+f(r)$ where $f(r)$ becomes small at large $r$ and $\mathrm df/\mathrm dr \propto 1/r$, but that would give $f = \ln(r)$ and that doesn't go to unity at large $r$. No doubt our more experienced mathematicians can immediately think of a solution, but I have to confess that nothing springs to mind.

John Rennie
  • 355,118
  • 2
    This is correct. And there's nothing wrong with your solution. It's spherically symmetric, but not asymptotically flat. It is also not a vacuum solution, which probably expalains why it's not asymptotically flat. Since it's not asymptotically flat, it's "ADM Mass" (quotes because ADM mass requires asymptotic flatness) is infinite. This is a physically reasonable universe, it's just in no way our universe. – Zo the Relativist Sep 01 '16 at 14:29
  • 3
    If in the limit $r\to\infty$ we have $b(r)\to 1$ then we must have $da(r)/dr\sim r^{-1}$ in which case $a(r)\sim \log(r)$ but no such function approaches unity as $r\to\infty$. So it seems to follow that the answer is 'no, there is no such metric'? – lemon Sep 01 '16 at 14:29
  • 1
    FWIW spherical symmetry is not a requirement. I guess I would stick to having some kind of rotational symmetry (perhaps cylindrical). Asymptotic flatness would be nice, but it's looking like that's not going to work out. I hadn't really considered these issues in detail when asking the question. – David Z Sep 01 '16 at 14:59
  • Could a function that is ln(r) most of the time but often enough dives down and becomes unity before becoming ln(r) again fulfill both criteria? – Emil Sep 01 '16 at 15:47