# Light cones tipping over

• MeJennifer
In summary: Radial_coordinate#Normal_coordinates radial coordinate...of a point in space. However, because the Schwarzschild radius is related to the gravitational force, r can also take on a negative value. In that case, the coordinate system would be said to be in the negative radial coordinate direction.In summary, the "light cones tipping over" is a metaphor for time travel in GR. The tipping over is an indication that observers can't escape back out past the horizon. However, this reversal and tipping over is a peculiarity of the coordinates used in GR, and is not a property of the metric itself. There is no compelling argument for calling this a change from one
MeJennifer
"Light cones tipping over"

A common phrase used to show alleged time travel solutions in GR.
Even a person like Kip Thorne uses it.

But my question is, is that an accurate representation of GR in strong gravitational fields?

The Schwarzschild metric expressed using the Eddington-Finkelstein coordinates show those "light cones tipping over", and eventually the radial and time coordinates reverse.

But this reversal, and even the tipping over seems to me a peculiarity of the choice of coordinates. It seems to me that it is assumed that there is a particular relationship between the radial and time coordinate.
That seems to be a rather liberal interpretation of the metric.

Afteral nothing in the Minkowskian metric of ds2 = dt2 - dr2 implies that if dt2 becomes negative it must be considered space and if dr2 becomes negative it must be considered time.

Last edited:
MeJennifer said:
A common phrase used to show alleged time travel solutions in GR.
Even a person like Kip Thorne uses it.

But my question is, is that an accurate representation of GR in strong gravitational fields?

The Schwarzschild metric expressed using the Eddington-Finkelstein coordinates show those "light cones tipping over", and eventually the radial and time coordinates reverse.

But this reversal, and even the tipping over seems to me a peculiarity of the choice of coordinates. It seems to me that it is assumed that there is a particular relationship between the radial and time coordinate.
That seems to be a rather liberal interpretation of the metric.

Afteral nothing in the Minkowskian metric of ds2 = dt2 - dr2 implies that if dt2 becomes negative it must be considered space and if dr2 becomes negative it must be considered time.

Minkowski space-time, at least according to one of my textbooks, is specifically defined as a 4-D real linear space. Neither dt2 nor dr2 can ever become negative while dr and dt are real numbers. That wouldn't apply to a complex linear space though.

http://casa.colorado.edu/~ajsh/schwp.html may have useful diagrams for this question. Physically, the tipping of the light cones [which are traced out by null geodesics (a coordinate invariant idea)] is an indication that timelike observers traveling across the horizon cannot escape back out beyond the horizon.

robphy said:
http://casa.colorado.edu/~ajsh/schwp.html may have useful diagrams for this question. Physically, the tipping of the light cones [which are traced out by null geodesics (a coordinate invariant idea)] is an indication that timelike observers traveling across the horizon cannot escape back out beyond the horizon.
Yeah the tipping just means the future part of the light cones point into a point in space (the black hole).

MeJennifer said:
if dt2 becomes negative it must be considered space and if dr2 becomes negative it must be considered time.

Yep, keep in mind that "r" and "t" are just coordinates, they don't necessarilly have any global meaning. In particular, t certainly isn't purely timelike for all observers. The original motivation for my current work is all the existing EFE solutions which don't yet have much physical interpretation.

cesiumfrog said:
Yep, keep in mind that "r" and "t" are just coordinates, they don't necessarilly have any global meaning. In particular, t certainly isn't purely timelike for all observers. The original motivation for my current work is all the existing EFE solutions which don't yet have much physical interpretation.
But even in coordinate space, what is the compelling argument for calling this a change from one to the other? Would that not imply that t and r are both related to each others imaginary plane? So that if t2 becomes negative it becomes space and that if r2 becomes negative it becomes time?
Would that not be a rather liberal interpretation of the Minkowski metric?

In plain Schwarzschild the light cones get narrower the closer one gets to the event horizon and eventually becomming zero. I would think it much more logical to see them get an imaginary size past the event horizon than to assume that, magically, the time and space coordinates get switched.

You say to keep in mind that coordinate space is not observer space is of course true. But, people use the tipping of light cones, which by the way is only happening when you use the Eddington-Finkelstein coordinates, to "explain" closed time curves. That again is a rather premature conclusion IMHO.

But on the other hand we see that people like Kip Thorne and more surprisingly Roger Penrose use it as well. Penrose who at one point proposed the use of a complex compactified Minkowski spacetime. Obviously this spacetime does not satisfy the condition that t and r are both related to each others imaginary plane.
So go figure, what's the real story here?

Last edited:
MeJennifer said:
..if r2 becomes negative..
Normally you would leave r as simply a real valued coordinate; it's the metric components (eg. $g_{rr}$ vs. $g_{tt}$) that may change in sign.

MeJennifer said:
In plain Schwarzschild the light cones get narrower the closer one gets to the event horizon and eventually becomming zero. ...
I don't know that it makes any sense to say that (by what additional metric are you measuring the cones?)..

In the plain old (ie. Boyer-Lindquist) coordinates, the light cones "tip" as you approach the event horizon, which means light rays begin to traverse greater intervals of r coordinate (note we're implicity comparing to a euclidean metric over the same chart). From points on the horizon (neglecting that the chart is actually invalid just there), the cone would be rotated such that the future cone is centered towards the negative r direction, and the past cone in the direction of increasing r coordinate.

MeJennifer said:
But, people use the tipping of light cones, which by the way is only happening when you use the Eddington-Finkelstein coordinates, to "explain" closed time curves.
The tipping happens in general, unless you specifically choose a coordinate system to avoid it. The relation to causality violation can occur for example if null cones in a coordinate space tip such that the $\phi$ coordinate becomes everywhere timelike AND if there is also some kind of closed geometry such that having zero $\phi$ coordinate be identified with having $\phi$ coordinate equal to $2\pi$.

cesiumfrog said:
I don't know that it makes any sense to say that (by what additional metric are you measuring the cones?)..
Using standard Schwarzschild coordinates light slows down for decreasing values of r. That is why the cones get narrower. But remember that r is nothing but a coordinate, it does not represent the physical distance from the center of gravity.

cesiumfrog said:
The tipping happens in general, unless you specifically choose a coordinate system to avoid it.
Using standard Schwarzschild coordinates light cones do not tip over.
They also do not tip using Kruskal coordinates. However they do tip over using the Eddington-Finkelstein coordinates.

Anyway we can do a lot of juggling with coordinates and then demonstrating all kinds of effects that are completely non physical but only due to the usage of certain coordinate systems.

More important is the critical review of the suggestion that time can become like space and space can become like time. I know for some that is a done deal and I simply do not "understand" it but to me it is not a done deal at all.

Last edited:
MeJennifer said:
But remember that r is nothing but a coordinate, it does not represent the physical distance from the center of gravity.

MeJennifer said:
Using standard Schwarzschild coordinates light cones do not tip over.
By saying "standard", you are referring to "Boyer-Lindquist", right?
In any case, say you draw a t vs. r plot (standard coords) depicting the null cones for observers at different positions.. would you not describe them as "tipping"?

MeJennifer said:
Anyway we can do a lot of juggling with coordinates and then demonstrating all kinds of effects that are completely non physical but only due to the usage of certain coordinate systems.

More important is the critical review of the suggestion that time can become like space and space can become like time. I know for some that is a done deal and I simply do not "understand" it but to me it is not a done deal at all.

Hmm. I would really avoid saying "time can become like space" or vice-versa, because I don't think it means anything. Physically, to an observer journeying across the event horizon, nothing ever happens of local significance. It's a confusing coincidence that the original Schwarzschild coordinates have undesirable features exactly where the event horizon is... Just that when we try to draw some set of coordinate axis lines (initially chosen somewhat arbitrarily), we sometimes find one of those lines is timelike in one region, and spacelike in another.

cesiumfrog said:
By saying "standard", you are referring to "Boyer-Lindquist", right?
Boyer-Lindquist is used for a rotating mass not for a non rotating mass.

Using Schwarzschild coordinates there is only one r coordinate. Eddington-Finkelstein introduces an additional r* called the tortoise coordinate. With this in place we get tilting cones.
And if one additional coordinate is not enough we can always use the Krusal coordinates that introduces two coordinates that relate to t and r. In Krusal coordinates the cones are always straight up and also 45 degrees (so no narrowing).

cesiumfrog said:
In any case, say you draw a t vs. r plot (standard coords) depicting the null cones for observers at different positions.. would you not describe them as "tipping"?
A very good question!
For starters it would not be applicable for Schwarzschild coordinates since this is a view of the situation from the perspective of a distant observer far removed from the gravitational field. So the point is that in Schwarzschild coordinates the observer is fixed.

But with regards to a observer in free fall who is approaching the event horizon and continuing towards the singularity, the speed of light will remain c, at least locally, so that implies that the angle of the cone would remain constant. However we can hardly speak of a cone from the observer's perspective, it would seem clear that since the curvature for this observer is so strong that a construction of an orthonormal coordinate system would show anything but a cone for incoming and outgoing light rays except for a very small local region. Past the event horizon the observer would still measure the speed of light at c, again only locally, but because the curvature is getting so strong here that the closure between the observer and the center of mass is faster than the width angle of the observer's light cone would allow. Now some seem to interpret this as time becoming space and vice versa, which is the whole point of this topic.

It's a confusing coincidence that the original Schwarzschild coordinates have undesirable features exactly where the event horizon is...
To me that is not a coincidence at all. Remember that the Schwarzschild model is a view from the perspective of an observer far removed from the gravitational influence.

Last edited:
MeJennifer said:
A very good question!
For starters it would not be applicable...

If so I disagree. Regardless of whether you use Schwarzschild or Ingoing (Eddington-Finkelstein) coordinates, the tilting is what you see when you draw light "cones" on a t vs. r (or t* vs. r*) plot. It's a coordinate effect.

Since coordinates are nothing physical, and you can't ever speak of shape or orientation of a light cone from the observer's perspective, why make a big deal of the tilting at all?

As for time travel, that occurs when the tilting is such that one's future light cone intersects one's past light cone. It isn't an issue in a black hole (the future cone is directed at the singularity, closed timelike loops don't become possible), it only means that from our perspective the gravity of a black hole sucks things away too strongly for their rockets or even their radio messages to approach us.

cesiumfrog said:
If so I disagree. Regardless of whether you use Schwarzschild or Ingoing (Eddington-Finkelstein) coordinates, the tilting is what you see when you draw light "cones" on a t vs. r (or t* vs. r*) plot. It's a coordinate effect.
Well I am quite sure that you would not accept from me that using Schwarzschild coordinates light cones do not tip, and that instead you need Eddington-Finkelstein coordinates for that.
Perhaps from someone else you will.

Last edited:
I accept I may be wrong (you haven't yet motivated me to draw the plot myself) but my reasoning is that in Schwarzschild coordinates: far from the black hole, light tends to travel in the time direction (the cone spreading equally in +/- radial directions); inside the even horizon light must always travel radially inward (presumably also spreading in the t coordinate directions); if anything Ingoing coordinates would demonstrate less tilt (limited to 45 degrees).

Regardless, if you believe the tilting appears only in one of two equally valid coordinate systems, do you agree that "space becoming time" is not a physical effect (in black holes)?

cesiumfrog said:
I accept I may be wrong (you haven't yet motivated me to draw the plot myself) but my reasoning is that in Schwarzschild coordinates: far from the black hole, light tends to travel in the time direction (the cone spreading equally in +/- radial directions); inside the even horizon light must always travel radially inward (presumably also spreading in the t coordinate directions); if anything Ingoing coordinates would demonstrate less tilt (limited to 45 degrees).
The whole issue IMHO with cones is that they work well with SR but miserably fail with GR.

In flat spacetime we have cones, but not in curved spacetime. Yes of course if we take a very small region that we can consider flat then we have a mini cone but in curved spacetime we cannot speak about a cone at all. Sure we can insist on cones by picking the proper coordinate system, but it does not mean anything, for all intents and purposes we might as well create a coordinate system that shows cones as passa doblé steps.

cesiumfrog said:
Regardless, if you believe the tilting appears only in one of two equally valid coordinate systems, do you agree that "space becoming time" is not a physical effect (in black holes)?
I agree that such coordinate representations are more or less meaningless from a physical perspective.

But the problem is that tipping of cones is shown as theoretical evidence for things like closed time loops. Even people like Kip Thorne, Roger Penrose and Stephen Hawking seem run away with it and write books that "clearly shows" what is going on. In my understanding at least it seem that they should know better, but clearly I just don't seem to understand why it is obvious that time and space can flip inside the event horizon. Any takers on a simple explanation?

The main issue that I wanted to bring up in this topic is that suggestion, the suggestion that time becomes like space and vice versa. I don't see any indication for that, except for when we make a particular interpretation of the time and space relationship in the Minkowski metric.

Last edited:
The interpretation is natural though since asymptotically that's the preffered coordinate system.

The choice of tipping is indeed a coordinate artifact, but it makes good sense from a physical standpoint to compare it with what we know best. Eg a rockets journey starts out with the obvious Minkowski coordinates system, and as you get close and pass through the black holes event horizon it serves as an illustration how either you have to abandon your choice of coordinates, or you have to admit that ones notion of time/space are going to warp and switch places.

And there's nothing wrong with writing down lightcones locally. Indeed we specifically choose not to pick coordinates where they are torus's or something like that, b/c no one has any intuition whatseover about that and indeed most calculational strategies vanish with such a stupid choice of local coordinates.

Haelfix said:
The choice of tipping is indeed a coordinate artifact, but it makes good sense from a physical standpoint to compare it with what we know best. Eg a rockets journey starts out with the obvious Minkowski coordinates system, and as you get close and pass through the black holes event horizon it serves as an illustration how either you have to abandon your choice of coordinates, or you have to admit that ones notion of time/space are going to warp and switch places.
So demonstrate to me how they switch place!
How do two separate dimensions get intertwined on a Riemann surface?

Last edited:
MeJennifer said:
The whole issue IMHO with cones is that they work well with SR but miserably fail with GR.

In flat spacetime we have cones, but not in curved spacetime. Yes of course if we take a very small region that we can consider flat then we have a mini cone but in curved spacetime we cannot speak about a cone at all. Sure we can insist on cones by picking the proper coordinate system, but it does not mean anything, for all intents and purposes we might as well create a coordinate system that shows cones as passa doblé steps.

In Minkowski spacetime M, we have two ways to think of the "light cone of an event p"...

It is the "set of events in M" that can only reach or be reached by the vertex event p by a light ray. It also forms the boundary between events that can be causally connected to p from those that cannot.

It is also a "set of directions [set of vectors] in TpM (the tangent space at p)" that are tangent to lightlike paths through event p (i.e., "set of lightlike tangent vectors at p"). It also forms the boundary between the spacelike and non-spacelike (i.e. causal) tangent vectors. TpM is a vector space with a Minkowski metric.

In a general curved spacetime, the notion of "light cone of an event p" usually means the "set of lightlike tangent vectors at p".

Nothing I have said above depends on any choice of coordinates.

To draw these light cones in the spacetime diagram of a spacetime, curved or otherwise, one must identify all of the lightlike geodesics (a coordinate invariant notion). At a particular event, its light cone is determined by the tangents to these geodesics.

Depending on your choice of coordinates, the image of these geodesics in your coordinate chart may trace out all sorts of crazy looking paths (akin to the distortions one gets from various map projections of the earth). In some cases, the image of these light cones may look tipped or distorted relative to the images of other light cones in your coordinate chart.

Regardless of appearances in your chart, the physics is determined by the lightlike geodesics, essentially telling you which events are in the causal future [and causal past] of events in spacetime (i.e. the causal connectivity of events). The worldlines of observers are bounded by these lightlike geodesics.

It may turn out that the null geodesics tell you that you might have closed causal curves. Or it may turn out that certain sets of events have causal futures that don't extend to spatial or null infinity. Or something other feature that one doesn't see in Minkowski spacetime.

Depending on your choice of chart (and thus the images of the light cones), it may be easier or harder to tell the story of what is going on physically. Depending on the particular aspect of the story you want to tell, some charts are better suited than others.

MeJennifer said:
But the problem is that tipping of cones is shown as theoretical evidence for things like closed time loops.
More correctly, these show that closed time loops are mathematically possible, given the constraints imposed in the situation. In other words, saying that one has a 4-manifold with a Lorentzian-signature metric places some restrictions on what "physics" is possible. However, by themselves, they don't restrict the possibility of closed time loops or other pathologies. Even imposing the field equations might still allow pathologies. That is why one is led to the notion of "causality conditions" and the study of "causal structure", which were developed using "global methods" (i.e. geometric, coordinate-free methods). One may also impose other conditions like "energy conditions", "asymptotic conditions", etc...

Last edited:
robphy said:
To draw these light cones in the spacetime diagram of a spacetime, curved or otherwise, one must identify all of the lightlike geodesics (a coordinate invariant notion). At a particular event, its light cone is determined by the tangents to these geodesics.
Sure I follow what you say here.

robphy said:
Depending on your choice of coordinates, the image of these geodesics in your coordinate chart may trace out all sorts of crazy looking paths (akin to the distortions one gets from various map projections of the earth). In some cases, the image of these light cones may look tipped or distorted relative to the images of other light cones in your coordinate chart.
Sure in certain coordinates.

robphy said:
Regardless of appearances in your chart, the physics is determined by the lightlike geodesics, essentially telling you which events are in the causal future [and causal past] of events in spacetime (i.e. the causal connectivity of events). The worldlines of observers are bounded by these lightlike geodesics.
Completely agree!

robphy said:
More correctly, these show that closed time loops are mathematically possible, given the constraints imposed in the situation. In other words, saying that one has a 4-manifold with a Lorentzian-signature metric places some restrictions on what "physics" is possible. However, by themselves, they don't restrict the possibility of closed time loops or other pathologies. Even imposing the field equations might still allow pathologies. That is why one is led to the notion of "causality conditions" and the study of "causal structure", which were developed using "global methods" (i.e. geometric, coordinate-free methods). One may also impose other conditions like "energy conditions", "asymptotic conditions", etc...
Well at one point the time part of the geodesic has to connect to another time part of the same geodesic while the spatial parts are irrelevant. Apart from a closed spacetime or a wormhole I do not see how that can be the case. Can you?

robphy said:
Please do, to me it makes absolutely no sense.

MeJennifer said:
So demonstrate to me how they switch place!
Ok you guys are much better than I am at this, but I thought I had a handle on it. Let's look at this in the Schwarzschild coordinates. Is it sufficient to show t becomes imaginary & that a decrease in r becomes as inevitable as going forward in time?

Last edited:
In case anyone's still interested in this, the clearest pedagogical explanation of this is (unsurprisingly) in MTW. Check out pp. 823-826 for a discussion that uses the Schwarzschild geometry and the surface at $$r=2M$$ as an example.

See this posting in the https://www.physicsforums.com/showpost.php?p=1146536&postcount=21" topic for a comment that I think is applicable to this topic as well.

George Jones said:
For a large, ideal non-rotationg black hole, which is a valid solution to the equations of GR, both our senses and GR work fine within the horizon. In particular, time and space don't "swap" inside the horizon. What does happen is that: a poor choice of labels is used; spacetime becomes non-stationary.

Here is an analogy. In some city, imagine that you are driving East on Bridge Street East. After the street makes a sharp left, you are driving almost north on Bridge Street East. East and North did not interchange, it is just that the labellng system has become poor.

In the same manner as street names are convenient labels that humans assign to positions in cities, spacetime coordinates (like r) are just labels assigned by humans to spacetime events. Inside the event horizon, r is a timelike coordinate, so it would make more sense to chanlge the name of the human-assigned label r to something more descriptive. For (partially) historical reasons, this isn't done.

Similarly, in the above city, after the left, it would make sense to change the human-assigned label Bridge Street East to something like Bridge Street North. This hasn't been (and won't be) done, because the city's inhabitants have been calling it Bridge Street East since before anyone can remember.

Last edited by a moderator:
Appearance of light cones in curved spacetimes

Hi again, Jennifer,

MeJennifer said:
A common phrase used to show alleged time travel solutions in GR. Even a person like Kip Thorne uses it.

But my question is, is that an accurate representation of GR in strong gravitational fields?

It is when Kip Thorne uses it! :-/ I know that because he can provide a correct figure which conforms to this informal description, as I can verify using my own computations.

I prefer to be more specific about this "tipping". I guess you are talking about light cones in the Boyer-Lindquist chart for the Kerr vacuum solution in gtr, which does feature closed timelike curves in the interior region, or light cones in the Goedel lambdadust solution, which also features closed timelike curves (see for example the beautiful figures in Hawking and Ellis, Large Scale Structure of Space-Time, for both of these examples).

But let's study an even simpler example:

MeJennifer said:
The Schwarzschild metric expressed using the Eddington-Finkelstein coordinates show those "light cones tipping over", and eventually the radial and time coordinates reverse.

Specifically, consider the advanced (infalling) Eddington chart, in which the line element takes the form
$ds^2 = -(1-2 m/r) \, du^2 + 2 \, du \, dr + r^2 \, \left( d\theta^2 + \sin(\theta) \, d\phi^2 \right),$
$-\infty < u < \infty, \; 0 < r < \infty, \; 0 < \theta < \pi, \; -\pi < \phi < \pi$
We can write down a "frame field" consisting of four orthonormal vector fields, a timelike unit vector
$\vec{e}_1 = \partial_u - m/r \, \partial_r$
plus three spacelike unit vectors
$\vec{e}_2 = \partial_u - (1-m/r) \, \partial_r$
$\vec{e}_3 = 1/r \, \partial_\theta$
$\vec{e}_4 = 1/r/\sin(\theta) \, \partial_\phi$
You can use these to draw the light cones. If you do it right, they will all be tangent to the null vector field $$\partial_r$$ and as r decreases, they lean inwards, until at $$r=2 m$$ they are also tangent to $$\partial_u$$.

MeJennifer said:
eventually the radial and time coordinates reverse.

Many people, even some who ought to know better, do talk that way, and invariably they wind up confusing everyone, including themselves. What they should really say is that the vectors $$\partial_u$$ are timelike outside the horizon, null at the horizon, and spacelike inside the horizon. Nothing "reverses"; in particular, the frame vectors given above are unambiguously timelike throughout (for the first) or spacelike throughout (for the remaining three).

George Jones is completely correct: of course "time and space" do not "swap roles" inside the horizon, that would be nonsense!

To elaborate on one point he alluded to, the coordinate basis vector field $$\partial_u$$ happens to be a Killing vector field; that is, the Schwarzschild vacuum is invariant under time translation. Similarly, the coordinate basis vector $$\partial_\phi$$ is a spacelike Killing vector whose integral curves are circles; that is, the Schwarzschild vacuum is invariant under rotation about the axis r=0.

The fact that in the exterior we have an irrotational timelike Killing vector and a spacelike Killing vector (whose integral curves are circles) means that the exterior region is static and axisymmetric. (This is also true of the Kerr vacuum solution.) Inside, we have two spacelike Killing vectors, but no timelike Killing vector; the solution is NOT static inside the horizon. Of course not, since otherwise an observer could use his rocket engine to hover at some Schwarzschild radius $$0 < r < 2m$$.

MeJennifer said:
But this reversal, and even the tipping over seems to me a peculiarity of the choice of coordinates. It seems to me that it is assumed that there is a particular relationship between the radial and time coordinate.

Not sure I understand that, but it sounds like you did correctly recognize that the coordinate basis vector $$\partial_u$$ changes character at the horizon.

Hope this helps,

Chris Hillman

Last edited:
Chris Hillman said:
...the Kerr vacuum solution in gtr, which does feature closed timelike curves in the interior region...

Could you elaborate a little on exactly where those closed timelike curves are in the Kerr solution?

CTC's in the Kerr vacuum

Hi, Cesium,

cesiumfrog said:
Could you elaborate a little on exactly where those closed timelike curves are in the Kerr solution?

The system still seems (as least to me) to be quite unstable (but then I've only been here for a few days), so I daren't try to write very much (having lost quite a bit of work here in the past few days), but briefly, the CTCs in the Kerr vacuum are all located in the "deep interior" blocks (referring to the usual Carter-Penrose block or conformal diagram). The no-hair theorems do not imply that the -interior- geometry prefers to be Kerr-like, and there are various considerations which suggest that it should not be, quite apart from our natural desire to avoid predicting CTCs even in places where, even if a physicist should experience such weirdness, he'd be unable to report this to his colleages in the exterior.

The Taub-NUT vacuum (Misner's "counterexample to everything") and Goedel lambdadust also exhibit some startling causal structure. In fact, the best example to become familiar with CTC's is probably the Goedel lambdadust solution. As it happens, I just came across a spanking new arXiv eprint which offers an extensive and well illustrated discussion; see http://www.arxiv.org/abs/gr-qc/0611093

Chris Hillman

Regarding the "tipping of lightcones" can someone point a coordinate system for Schwarzschild black hole, where this does not happen? At least with the usual suspects this seems to happen ("the nature of dx -> dt" in the usual way or the relationship of the light cones and the tangent of fwo-path in Kruskal coordinates).

cesiumfrog said:
Could you elaborate a little on exactly where those closed timelike curves are in the Kerr solution?

Let me elaborate a bit on what Chris said.

O'Neill, in his book The Geometry of Kerr Black Holes, proves:

there is a closed timelike curve through any event inside the inner (Cauchy) horizon, i.e., through any event for which r < r-.

Carroll gives the following simple example. Consider a curve for which $\phi$ varies, and for which $t$, $r$, $\theta$ are held constant. Because of periodicity with respect to $\phi$, any such curve is closed.

Now, the timelike part.

Take $r < 0$ with $|r|$ small, and $\theta = \pi/2$. Note $r$ is a coordinate, not a radial distance, and negative $r$ is part of (extended) Kerr. Because $0 = dt = dr = d \theta$, the line element along the curve is

$$ds^2 = \left( r^2 + a^2 + \frac{2Mr a^2}{r^2} \right) d\phi^2$$

For $r$ negative and small. the last term, whcih is negative, dominates, and thus $ds^2$ is the line element for a timilike curve.

Last edited:
Utility of light cones, plus a Vaidya thought experiment

Hi, Jennifer,

MeJennifer said:
The whole issue IMHO with cones is that they work well with SR but miserably fail with GR.

I am coming into this thread very late, so this might not do much good, but I would say that to the contrary, drawing "infinitesimal" light cones in some coordinate chart is one of the most important things you can do when you are trying to understand the local geometry.

As so often happens in this subject, there is room for confusion here since some authors also use "light cone" (which when used without qualification generally means the concept defined at the level of a tangent space) to refer to the "absolute future" and "absolute past" of some event. For example, in discussions of Cauchy horizons or cosmological horizons.

MeJennifer said:
But the problem is that tipping of cones is shown as theoretical evidence for things like closed time loops. Even people like Kip Thorne, Roger Penrose and Stephen Hawking seem run away with it and write books that "clearly shows" what is going on.

I think it DOES show clearly what is going on, and I can't imagine how one could understand gtr without appealing to this kind of imagery.

MeJennifer said:
In my understanding at least it seem that they should know better, but clearly I just don't seem to understand why it is obvious that time and space can flip inside the event horizon.

Ah! I think this the key misconception, and I entirely agree with you that anyone who says "time and space can flip inside the event horizon" either doesn't understand how gtr treats the notion of a black hole, or else is being terribly negligent in attempting to give an informal verbal description of the mathematics. Forgive me if you already mentioned this somewhere, but can I ask what book(s) you are reading? If one of them is MTW, I hope that you will soon come to appreciate why anyone saying that "time and space flip roles inside a black hole" is speaking nonsense. On the other hand, if we use a coordinate chart, such as the Eddington chart
$ds^2 = -(1-2m/r) \, du^2 + 2 \, du \, dr + r^2 \left( d\theta^2 + \sin(\theta)^2 \, d\phi^2 \right),$
$-\infty < u < \infty, \; 0 < r < \infty, \; 0 < \theta < \pi, \; -\pi < \phi < \pi$
then the coordinate basis vector field $$\partial_u$$ is timelike outside the horizon but null on the horizon and spacelike inside the horizon.

Since $$\partial_u$$ is also a Killing vector field, the fact that it is null on the horizon means that in the Schwarzschild vacuum solution, the event horizon happens to be a "Killing horizon", i.e. it has a local characterization (it is the locus where our Killing vector field is null). But in general, the event horizon has no such description. A very similar model which slightly generalizes this is the Vaidya null dust (to obtain this, just consider m to be a monotonically increasing function of u in the Eddington chart above!) is a very good model to study in detail in order to understand the "teleological" global nature of the event horizon. The Vaidya null dust is not static (if m is increasing with u), $$\partial_u$$ is no longer a Killing vector field. In addition, if we allow a spherical shell of incoherent radiation to fall into the hole, thus increasing its mass and Schwarzschild radius, the event horizon smoothly increases, and the apparent horizon is not consistently defined during this increase. Even more startling, if the shell of incoming radiation carries a sufficient amount of energy, an observer hovering just outside the horizon could actually be inside the event horizon even BEFORE the radiation reaches him. Since it is moving at the speed of light, he cannot possibly obtain warning in time to hastily increase his distance from the hole. See the discussion in Frolov and Novikov, Black Hole Physics, for more detail plus some nice if microscopic pictures.

Just one more reason why the immediate vicinity of a black hole is probably not a very safe place to visit!

Chris Hillman

Last edited:
Chris Hillman said:
I am coming into this thread very late, so this might not do much good, but I would say that to the contrary, drawing "infinitesimal" light cones in some coordinate chart is one of the most important things you can do when you are trying to understand the local geometry.

As so often happens in this subject, there is room for confusion here since some authors also use "light cone" (which when used without qualification generally means the concept defined at the level of a tangent space) to refer to the "absolute future" and "absolute past" of some event. For example, in discussions of Cauchy horizons or cosmological horizons.
Well but a cone drawn at one particular point in curved space-time does not represent the complete past or the future since in GR past and future events are related within the confines of space-time curvature, clearly a picture that is hardly resembling anything close to a conical shape.

Chris Hillman said:
I think it DOES show clearly what is going on, and I can't imagine how one could understand gtr without appealing to this kind of imagery.
Well to me, but surely I am mistaken, a cone drawn at a particular point in curved space-time shows its past and future connections assuming that space-time is flat. Since space-time is curved the shape of the "cone" is anything but a cone.

Chris Hillman said:
Ah! I think this the key misconception, and I entirely agree with you that anyone who says "time and space can flip inside the event horizon" either doesn't understand how gtr treats the notion of a black hole, or else is being terribly negligent in attempting to give an informal verbal description of the mathematics.
Well I am glad I find someone who can relate to this.

But then, given that you agree with me they can't flip, how could some people even argue that some given space-time can form a closed time loop?

Chris Hillman said:
In addition, if we allow a spherical shell of incoherent radiation to fall into the hole, thus increasing its mass and Schwarzschild radius, the event horizon smoothly increases, and the apparent horizon is not consistently defined during this increase.
True, but then the Schwarzschild solution would no longer apply since the space-time is no longer static right?

CTC's again

Hi, Jennifer

MeJennifer said:
a cone drawn at one particular point in curved space-time does not represent the complete past or the future since in GR past and future events are related within the confines of space-time curvature, clearly a picture that is hardly resembling anything close to a conical shape.

That is actually one of the points I was hinting at. Of course, this doesn't mean that the absolute future of some event in a curved spacetime is not a valuable concept, even when (as in pp-wave models, for example) this concept breaks down globally.

MeJennifer said:
But then, given that you agree with me they can't flip, how could some people even argue that some given space-time can form a closed time loop?

You mean closed timelike curves (CTCs), which is something different (but sharing the same basic idea): in some Lorentzian manifolds, at least in some regions there exist closed timelike curves. For example, suppose we just solved the EFE to obtain the metric tensor written in something like cylindrical coordinates notices.

In fact, for concreteness let's take a specific example, the van Stockum dust solution (1937), which was historically the first solution in which CTCs were noticed, by Van Stockum (see the article "Willem Jakob van Stockum at http://en.wikipedia.org/wiki/User:Hillman/Archive#Contributors_to_general_relativity)
$ds^2 = -\left( dt -a\, r^2 d\phi \right)^2 + \exp(-a^2 \, r^2/2) \, \left( dz^2 + dr^2 \right) + r^2 \, d\phi^2,$
$-\infty < t < \infty, \; 0 < r < \infty, \; 0 < \theta < \pi, \; -\pi < \phi < \pi$
Multiplying out, we find that $$g_{\phi \phi} = \left( 1-a^2 \, r^2 \right) \, r^2$$, so that the coordinate basis vector $$\partial_\phi$$ is spacelike on $$0< r < 1/a$$, null at $$r=1/a$$, and timelike on $$1/a < r < \infty$$. This means that the circles $$t=t_0, \; z=z_0, \; r=r_0$$ are spacelike curves when $$r_0 < 1/a$$ but TIMELIKE when $$r_0 > 1/a$$.

At this point, you might wish to jump to the article "van Stockum dust" archived at http://en.wikipedia.org/wiki/User:Hillman/Archive#Contributors_to_general_relativity
Note in particular the pictures of how the (infinitesimal) light cones change in appearance as you increase radial coordinate in this spacetime.

Chris Hillman

Last edited:
Thanks Chris you gave me something to digest.

Chris Hillman said:
Multiplying out, we find that $$g_{\phi \phi} = \left( 1-a^2 \, r^2 \right) \, r^2$$, so that the coordinate basis vector $$\partial_\phi$$ is spacelike on $$0< r < 1/a$$, null at $$r=1/a$$, and timelike on $$1/a < r < \infty$$. This means that the circles $$t=t_0, \; z=z_0, \; r=r_0$$ are spacelike curves when $$r_0 < 1/a$$ but TIMELIKE when $$r_0 > 1/a$$.
Well, unquestionably, if we define the term 'timelike' as any coordinate that contributes positively (assuming the + - - - convention) to g it follows that this would indeed be a closed timelike loop.

But does this warrant a physical interpretation or, even stronger, a conclusion that a test mass traveling on this CTC meets its own past?

Last edited:
MeJennifer said:
Well, unquestionably, if we define the term 'timelike' as any coordinate that contributes positively (assuming the + - - - convention) to g it follows that this would indeed be a closed timelike loop.

But does this warrant a physical interpretation or, even stronger, a conclusion that a test mass traveling on this CTC meets its own past?
Certainly not.
It's just funny of mathematical model used (rather abusive one).
Make an experiment. That's the only way you will know for sure .

I wonder who first interpreted this "tipping of light cone" and r becomes time.
Was it Schwarzschild himself?

MeJennifer said:
Well, unquestionably, if we define the term 'timelike' as any coordinate that contributes positively (assuming the + - - - convention) to g it follows that this would indeed be a closed timelike loop.

But does this warrant a physical interpretation or, even stronger, a conclusion that a test mass traveling on this CTC meets its own past?

Given that the proper time experienced by an observer moving along a path is equal equal to the intergal of the Lorentz interval along that path (using a +--- sign convention), how could a closed path in space-time that was everywhere timelike not warrant such a physical interpretation?

pervect said:
Given that the proper time experienced by an observer moving along a path is equal equal to the intergal of the Lorentz interval along that path (using a +--- sign convention), how could a closed path in space-time that was everywhere timelike not warrant such a physical interpretation?
Just because a dimension d1 is orthogonal to a dimension d2 does not imply that d1 lies in d2's complex plane. It could be, but only if defined as such.

Feel free to demonstrate why Lorentz invariance implies such a connection.

Last edited:

• Special and General Relativity
Replies
4
Views
1K
• Special and General Relativity
Replies
10
Views
3K
• Special and General Relativity
Replies
8
Views
1K
• Special and General Relativity
Replies
13
Views
3K
• Special and General Relativity
Replies
1
Views
1K
• Special and General Relativity
Replies
3
Views
1K
• Special and General Relativity
Replies
11
Views
446
• Special and General Relativity
Replies
29
Views
1K
• Special and General Relativity
Replies
58
Views
4K
• Quantum Physics
Replies
10
Views
1K