We know that space-time dimensions are 3+1 macroscopically, but what if 2+2? Obviously it is tough to imagine two time dimensions, but mathematically we can always imagine as either having two parameters $t_1$ and $t_2$ or else in Lorentz matrix $$\eta_{00} = \eta_{11} = -1$$ and, $$\eta_{22} = \eta_{33} = 1.$$

Is there any physical reason for not taking this, like the norms become negative or something else?

  • 2
    $\begingroup$ Maybe you are interested the answer to this question too. $\endgroup$
    – Dilaton
    Nov 7, 2012 at 14:41
  • 1
    $\begingroup$ The "norms" could become negative anyway with usual Lorentzian signature. $\endgroup$
    – c.p.
    Nov 8, 2012 at 18:23
  • 3
    $\begingroup$ You may want to look at the following paper arxiv.org/abs/hep-ph/9910207 $\endgroup$
    – Newman
    Nov 9, 2012 at 2:20
  • $\begingroup$ About why there can not be more than one time dimension (I do not know how serious) space.mit.edu/home/tegmark/dimensions.pdf $\endgroup$
    – user126422
    Jun 22, 2017 at 4:06
  • $\begingroup$ In 'simulation theory' (yes yes... Wait for it) there would of course be a dimension of time in which our dimension is simulated. What I'm keen to know is... Does the two-time theory make predictions which are testable? Is there a way in which we could possibly know if we're in a simulation? And would this extra time dimension behave as the imagined time dimension of our simulators world would? $\endgroup$
    – Richard
    Jan 18, 2019 at 1:20

6 Answers 6


As Cumrun Vafa explains in the video linked to below the picture of him in this article, F-theory works in a total of $10+2$ dimensions. The signature of the last two infinitesimal dimensions is ambiguous, so that they can indeed both be timelike. Since these are only infinitesimal dimensions, any causality issues etc are not a problem in this case.

And as Cumrun Vafa nicely explains in his talk, F-theory gives quite a nice phenomenology with an astonishingly realistic CKM-Matrix, coupling constants, etc; so it is NOT true that theories that operate in more than one time dimension are completely off base, as some people claim. There is no reason to dogmatically dismiss every theory that has more than one time dimension.

BTW, the talk is very accessible and enjoyable.

  • $\begingroup$ This is true only if to consider any dimension with positive signature a time dimension. Which is wrong (in Newtonian spacetime, for instance, all dimensions have positive metric signature). $\endgroup$
    – Anixx
    Apr 25, 2021 at 2:19

The late Irving Segal of MIT had a theory where the usual Lorentz group was replaced by SO(4,2) and there were indeed two time dimensions. His book Mathematical Cosmology and Extragalactic Astronomy, Academic Press, 1976, worked out the details. His theory has not been generally accepted, although there may be a few mathematical physicists at Montreal who are still interested in it. One of the consequences of this "chronometry" as he called it was that a part of the observed redshift was merely due to the discrepancies between the two times, and was not a Doppler effect, and thus the universe was not expanding. This theory is not currently accepted.

He was a brilliant mathematician. He understood Physics. He did not understand how to do Physics. He made some great contributions to Mathematical Physics in his theorems about operator algebras, and those theorems were motivated by Physics. In fact, he was only interested in maths that was motivated by Physics.


The hyperbolicity of the associated classical field equations is lost in $d$ space plus $2$ time dimensions. One cannot define a locally SO(d,2)-invariant distinction between past an future, no matter how curled up one of the time dimensions is.

As a result, there is no way to implement causality (i.e., no way to enforce the limiting information transmission to a finite speed), and the resulting models have very little to do with the real world.

  • $\begingroup$ Well, 2+2 dimensions is obviously not compatible with reality, since we observe 3 spacelike dimensions. The paper by Dvali (see the comment by Newman) discusses the possibility that a timelike dimension is curled up. It's far from obvious to me that our universe couldn't act like it had hyperbolicity if it was 3+2-dimensional, but with one of the timelike dimensions curled up. $\endgroup$
    – user4552
    Apr 15, 2013 at 3:21
  • $\begingroup$ @BenCrowell: The concept of hyperbolicity is by definition tied to the Lorentzian metric! Curling up the second time direction does not change the signature of the metric. Thus even a curled up 2D time makes the distinction between spacelike and timelike impossible. $\endgroup$ Apr 22, 2015 at 9:10
  • $\begingroup$ the point being: that near-hyperbolicity should be recoverable from a 3+2 signature if one of the time dimensions is curled. If not, why not? $\endgroup$
    – lurscher
    Apr 29, 2019 at 22:18
  • $\begingroup$ @lurscher: In 3+2 dimensions, one cannot define a locally SO(3,2)-invariant distinction between past an future, no matter how curled up the time. $\endgroup$ Apr 30, 2019 at 9:09

There could actually be two time dimensions, though you have to implement those properly. Should you simply replace the domain of time from $\mathbb R$ to $\mathbb R^2$ you would completely mess up our notion of cause and effect, therefore giving rise to many logical and temporal paradoxes hard to explain. I don't even know if such a theory would be logically coherent.

However, these two time dimensions need not to belong to the same mathematical domain, nor to have an equivalent meaning. Maybe this link can be useful. See also this short review on wikipedia.


Although space-like and time-like coordinates enter in the same foot on a relativistic theory, there are physical differences between them, for example

  • One can travel back and forward through space-like coordinated, while it is impossible in time-like coordinates.

  • If there exist more than one time-like dimensions, that would mean our time is a linear combination of those. Since one does not see other time-like coordinated, it implies that the other (transverse) time-like coordinates are compact.

The presence of closed time-like coordinates spoils causality... reason why no more than a single time coordinate is considered in physics.


  • 2
    $\begingroup$ I do not agree. With a time cylinder model, with $T$ being the (constant) radius of the cylinder, the space-time interval would be (here $c = 1$) : $(\Delta s)² = (\Delta t)² + T (\Delta \theta)² - (\Delta r)²$. If $T$ is of the same order as the Planck time, the apparent violation of causality (in r and t) is only appreciable for a length separation of the same order as the Planck length. For standard lengths, the term $T (\Delta \theta)²$ is neglectable. $\endgroup$
    – Trimok
    Nov 14, 2012 at 9:47

The causality issues relevant to including a second dimension of time owe their incoherence to the controlling assumption that space is contiguous and time is continuous. A formulation that understands two dimensions of space and two of time works well when the spatial dimensions are considered separated by time and the temporal dimensions are separated by space. The resulting construct gives rise to a phenomenology of a continuous form of time and a contiguous form of space. Causality is not violated as causality is a phenomenological issue.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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