Why is general relativity in (2+1) dimensions different from cylindrical systems in (3+1) dimensional GR? The gravitational potential $\Phi$ of an infinite rod in newtonian gravity is $\Phi \sim \ln(r)$. This is the same as the gravitational potential of a point charge in two-dimensional Newtonian gravity (see https://en.wikipedia.org/wiki/Newtonian_potential). They are the same, because both systems exhibit cylindrical symmetry and Gauss Law yields a logarithmic potential in this case.
In general relativity the solution for a static cylindrical spacetime is the Levi-Civita spacetime, which in the Newtonian limit will also give a potential $\Phi \sim \ln(r)$ (see e.g. https://arxiv.org/abs/1901.06561).
But what I can't understand is that in (2+1) dimensional general relativity it is said the spacetime is flat outside of a mass point, so in the Newtonian limit $\Phi \sim 0$. This is claimed despite authors stating that a point particle in general relativity in (2+1) dimensions is equivalent to a (infinite) string in (3+1) dimensions.

"We discussed the global properties of the (locally flat) geometries generated by moving point particles in 2+1 dimensions, or equivalently by parallel moving cosmic strings in 3+1 dimensions." (Deser, Jackiw, t'Hooft (1992))


"There is also a close relation to cosmic strings in four dimensions since the space-time of an infinite straight string is effectively three-dimensional."  (Deser, Jackiw (1988))

So why is there a difference between Newtonian Potential derived from Levi-Civita spacetime and the Newtonian potential derived from a (2+1) dimensional general relativity? Is general relativity in (2+1) dimensions not simply a cross section through a cylindrical symmetric spacetime in (3+1) dimensions? What is it then?
 A: That's a good question! Let us consider vacuum solutions to the EFE with zero cosmological constant $\Lambda=0$:

*

*The static axisymmetric/cylindrical spacetimes in 3+1D are classified by Levi-Civita solutions [1,2]. They represent a (possibly thick$^1$) cosmic string of total zero angular momentum $J=0$. There are effectively 2 parameters:

*

*$\sigma$, which has an interpretation of an energy-density per length of the string.


*A deficit angle.
If there is no deficit angle, then the well-known Newtonian limit (with an attractive Newtonian $ln$ potential) exists and corresponds to the limit $\sigma\to 0^+$.


*The static spherical symmetric spacetimes in 2+1D. The corresponding Schwarzschild-like$^2$ solution is locally on Minkowski-form$^3$, and it has only 1 topological parameter: A deficit angle, which grows with the mass of the point source.
In both cases, if there is a deficit angle, it extends all the way to $r=\infty$, i.e. there is no asymptotic region $r=\infty$ unaffected by the string/point source, and hence there is no Newtonian limit.
Perhaps surprisingly, in 2+1D the Newtonian $ln$ potential has no analog in GR!
References:

*

*M.F.A. da Silva, L. Herrera, F.M. Paiva, & N.O. Santos, The Levi-Civita spacetime, arXiv:gr-qc/9607058
.


*K. Bronnikov, N.O. Santos, & Anzhong Wang, Cylindrical Systems in General Relativity, arXiv:1901.06561
.
--
$^1$ Angular motion of the thick cosmic string is not completely prohibited, which leads to 1 extra parameter, that saves the $\ln$ potential. This is the main answer to OP's title question.
$^2$ Not a black hole!
$^3$ This is related to the fact that GR in 2+1D is a topological field theory (TFT).
