According to this article on the European Space Agency web site just after the Big Bang and before inflation the currently observable universe was the size of a coin. One millionth of a second later the universe was the size of the Solar System, which is an expansion much much faster than speed of light. Can space expand with unlimited speed?

  • 2
    $\begingroup$ Related: physics.stackexchange.com/q/20056 $\endgroup$ Commented Apr 9, 2013 at 11:39
  • 1
    $\begingroup$ More related: physics.stackexchange.com/q/44386 $\endgroup$
    – user10851
    Commented May 17, 2013 at 23:48
  • 1
    $\begingroup$ Possible duplicates: physics.stackexchange.com/q/26549/2451 and links therein. $\endgroup$
    – Qmechanic
    Commented May 22, 2013 at 20:31
  • 7
    $\begingroup$ The expansion hasn't got a speed. It is a misnomer to say that it is a speed. It should be called expansion rate. It is not like two points having a relative speed, it is more like a scaling rate of the unit distance. If there were no masses in the universe we would not sense any expansion at all. $\endgroup$ Commented Apr 15, 2018 at 8:48
  • 1
    $\begingroup$ In "Many Worlds In One", Vilenkin remarks that expansion differs from relative motion. However, as quoted in their preprint whose main title was "Expanding Confusion", Davis (of Lineweaver and Davis, cited in Pulsar's answer) had rejected the physicality of "spatial expansion" as any sort of force or drag. I'm making this comment mainly to help myself find Oktay Dogangun's helpful comment & Pulsar's answer, whose color-coded diagram is easier to follow than a couple of the 3 panels in Lineweaver & Davis's paper. $\endgroup$
    – Edouard
    Commented Oct 9, 2021 at 3:01

5 Answers 5


There are quite a few common misconceptions about the expansion of the universe, even among professional physicists. I will try to clarify a few of these issues; for more information, I highly recommend the article "Expanding Confusion: common misconceptions of cosmological horizons and the superluminal expansion of the Universe" from Tamara M. Davis and Charles H. Lineweaver.

I will assume a standard ΛCDM-model, with $$ \begin{align} H_0 &= 67.3\;\text{km}\,\text{s}^{-1}\text{Mpc}^{-1},\\ \Omega_{R,0} &= 9.24\times 10^{-5},\\ \Omega_{M,0} &= 0.315,\\ \Omega_{\Lambda,0} &= 0.685,\\ \Omega_{K,0} &= 1 - \Omega_{R,0} - \Omega_{M,0} - \Omega_{\Lambda,0} = 0. \end{align} $$

The expansion of the universe can be described by a scale factor $a(t)$, which can be thought of as the length of an imaginary ruler that expands along with the universe, relative to the present day, i.e. $a(t_0)=1$ where $t_0$ is the present age of the universe.

From the standard equations, one can derive the Hubble parameter $$ H(a) = \frac{\dot{a}}{a} = H_0\sqrt{\Omega_{R,0}\,a^{-4} + \Omega_{M,0}\,a^{-3} + \Omega_{K,0}\,a^{-2} + \Omega_{\Lambda,0}}, $$ such that $H(1)=H_0$ is the Hubble constant. In a previous post, I showed that the age of the universe, as a function of $a$, is $$ t(a) = \frac{1}{H_0}\int_0^a\frac{a'\,\text{d}a'}{\sqrt{\Omega_{R,0} + \Omega_{M,0}\,a' + \Omega_{K,0}\,a'^2 + \Omega_{\Lambda,0}\,a'^4}}, $$ which can be numerically inverted to yield $a(t)$, and consequently $H(t)$. It also follows that the present age of the universe is $t_0=t(1)=13.8$ billion years.

Now, another consequence of the Big Bang models is Hubble's Law, $$ v_\text{rec}(t_\text{ob}) = H(t_\text{ob})\,D(t_\text{ob}), $$ describing the relation between the recession velocity $v_\text{rec}(t_\text{ob})$ of a light source and its proper distance $D(t_\text{ob})$, at a time $t_\text{ob}$. In fact, this follows immediately from the definition of $H(t_\text{ob})$, since $v_\text{rec}(t_\text{ob})$ is proportional to $\dot{a}$ and $D(t_\text{ob})$ is proportional to $a$.

However, it should be noted that this is a theoretical relation: neither $v_\text{rec}(t_\text{ob})$ nor $D(t_\text{ob})$ can be observed directly. The recession velocity is not a "true" velocity, in the sense that it is not an actual motion in a local inertial frame; clusters of galaxies are locally at rest. The distance between them increases as the universe expands, which can be expressed as $v_\text{rec}(t_\text{ob})$. Some cosmologists therefore prefer to think of $v_\text{rec}(t_\text{ob})$ as an apparent velocity, a theoretical quantity with little physical meaning.

A related quantity that is observable is the redshift of a light source, which is the cumulative increase in wavelength of the photons as they travel through the expanding space between source and observer. There is a simple relation between the scale factor and the redshift of a source, observed at a time $t_\text{ob}$: $$ 1 + z(t_\text{ob}) = \frac{a(t_\text{ob})}{a(t_\text{em})}, $$ such that the observed redshift of a photon immediately gives the time $t_\text{em}$ at which the photon was emitted.

The proper distance $D(t_\text{ob})$ of a source is also a theoretical quantity. It's an "instantaneous" distance, which can be thought of as the distance you would obtain with a (very long!) measuring tape if you were able to "stop" the expansion of the universe. It can however be derived from observable quantities, such as the luminosity distance or the angular diameter distance. The proper distance to a source, observed at time $t_\text{ob}$ with a redshift $z_\text{ob}$ is $$ D(z_\text{ob},t_\text{ob}) = a_\text{ob}\frac{c}{H_0}\int_{a_\text{ob}/(1+z_\text{ob})}^{a_\text{ob}}\frac{\text{d}a}{\sqrt{\Omega_{R,0} + \Omega_{M,0}\,a + \Omega_{K,0}\,a^2 + \Omega_{\Lambda,0}\,a^4}}, $$ with $a_\text{ob} = a(t_\text{ob})$. The furthest objects that we theoretically can observe have infinite redshift; they mark the edge of the observable universe, also known as the particle horizon. Ignoring inflation, we get: $$ D_\text{ph}(t_\text{ob}) = a_\text{ob}\frac{c}{H_0}\int_0^{a_\text{ob}}\frac{\text{d}a}{\sqrt{\Omega_{R,0} + \Omega_{M,0}\,a + \Omega_{K,0}\,a^2 + \Omega_{\Lambda,0}\,a^4}}. $$ In practice though, the furthest we can see is the CMB, which has a current redshift $z_\text{CMB}(t_0)\approx 1090$.

A source that has a recession velocity $v_\text{rec}(t_\text{ob})=c$ has a corresponding distance $$ D_\text{H}(t_\text{ob})=\frac{c}{H(t_\text{ob})}. $$ This is called the Hubble distance.

Almost there, just a few more quantities need to be defined. The photons that we observe at a time $t_\text{ob}$ have travelled on a null geodesic called the past light cone. It can be defined as the proper distance that a light source had at a time $t_\text{em}$ when it emitted the photons that we observe at $t_\text{ob}$: $$ D_\text{lc}(t_\text{em},t_\text{ob})= a_\text{em}\frac{c}{H_0}\int_{a_\text{em}}^{a_\text{ob}}\frac{\text{d}a}{\sqrt{\Omega_{R,0} + \Omega_{M,0}\,a + \Omega_{K,0}\,a^2 + \Omega_{\Lambda,0}\,a^4}}. $$ There are two special cases: for $t_\text{ob}=t_0$ we have our present-day past light cone (i.e. the photons that we are observing right now), and for $t_\text{ob}=\infty$ we get the so-called cosmic event horizon: $$ D_\text{eh}(t_\text{em})= a_\text{em}\frac{c}{H_0}\int_{a_\text{em}}^\infty\frac{\text{d}a}{\sqrt{\Omega_{R,0} + \Omega_{M,0}\,a + \Omega_{K,0}\,a^2 + \Omega_{\Lambda,0}\,a^4}}. $$ For light emitted today, $t_\text{em}=t_0$, this has a special significance: if a source closer to us than $D_\text{eh}(t_0)$ emits photons today, then we will be able to observe those at some point in the future. In contrast, we will never observe photons emitted today by sources further than $D_\text{eh}(t_0)$.

One final definition: instead of proper distances, we can use co-moving distances. These are distances defined in a co-ordinate system that expands with the universe. In other words, the co-moving distance of a source that moves away from us along with the Hubble flow, remains constant. The relation between co-moving and proper distance is simply $$ D_c(t) = \frac{D(t)}{a(t)}, $$ so that both are the same at the present day $a(t_0)=1$. Thus $$ \begin{align} D_\text{c,ph}(t_\text{ob}) &= \frac{D_\text{ph}(t_\text{ob})}{a_\text{ob}},\\ D_\text{c,lc}(t_\text{em},t_\text{ob}) &= \frac{D_\text{lc}(t_\text{em},t_\text{ob})}{a_\text{em}},\\ D_\text{c,H}(t_\text{ob}) &= \frac{D_\text{H}(t_\text{ob})}{a_\text{ob}}. \end{align} $$ In fact, it would have been more convenient to start with co-moving distances instead of proper distances; in case you've been wondering where all the above integrals come from, those can be derived from the null geodesic of the FLRW metric: $$ 0 = c^2\text{d}t^2 - a^2(t)\text{d}\ell^2, $$ such that $$ \text{d}\ell = \frac{c\,\text{d}t}{a(t)} = \frac{c\,\text{d}a}{a\,\dot{a}} = \frac{c\,\text{d}a}{a^2\,H(a)}, $$ and $\text{d}\ell$ is the infinitesimal co-moving distance.

So, what can we do with all these tedious calculations? Well, we can draw a graph of the evolution of the expanding universe (after inflation). Inspired by a similar plot in the article from Davis & Lineweaver, I made the following diagram:enter image description here

This graph contains a lot of information. On the horizontal axis, we have the co-moving distance of light sources, in Gigalightyears (bottom) and the corresponding Gigaparsecs (top). The vertical axis shows the age of the universe (left) and the corresponding scale factor $a$ (right). The horizontal thick black line marks the current age of the universe (13.8 billion years). Co-moving sources have a constant co-moving distance, so that their world lines are vertical lines (the black dotted lines correspond with sources at 10, 20, 30, etc Gly). Of course, our own world line is the thick black vertical line, and we are currently situated at the intersection of the horizontal and vertical black line.

The yellow lines are null geodesics, i.e. the paths of photons. The scale of the time axis is such that these photon paths are straight lines at 45° angles. The orange line is our current past light cone. This is the cross-section of the universe that we currently observe: all the photons that we receive now have travelled on this path. The path extends to the orange dashed line, which is our future light cone. The particle horizon, i.e. the edge of our observable universe, is given by the blue line; note that this is also a null geodesic. The red line is our event horizon: photons emitted outside the event horizon will never reach us.

The purple dashed curves are distances corresponding with particular redshift values $z(t_\text{ob})$, in particular $z(t_\text{ob}) = 1, 3, 10, 50, 1000$. Finally, the green curves are lines of constant recession velocity, in particular $v_\text{rec}(t_\text{ob}) = c, 2c, 3c, 4c$. Of course, the curve $v_\text{rec}(t_\text{ob}) = c$ is nothing else than the Hubble distance.

What can we learn from all this? Quite a lot:

  • The current (co-moving) distance of the edge of the observable universe is 46.2 billion ly. Of course, the total universe can be much bigger, and is possibly infinite. The observable universe will keep expanding to a finite maximum co-moving distance at cosmic time $t = \infty$, which is 62.9 billion ly. We will never observe any source located beyond that distance.
  • Curves of constant recession velocity expand to a maximum co-moving distance, at $t_\text{acc} = 7.7$ billion years, and then converge again. This time $t_\text{acc}$, indicated by the horizontal black dashed line, is in fact the moment at which the expansion of the universe began to accelerate.
  • Curves of constant redshift also expand first, and converge when $t$ becomes very large. This means that a given source, which moves along a vertical line, will be observed with an infinite redshift when it enters the particle horizon, after which its redshift will decrease to a mimimum value, and finally increase again to infinity at $t = \infty$. In other words, every galaxy outside our local cluster will eventually be redshifted to infinity when the universe becomes very old. This is due to the dominance of dark energy at late cosmic times. Photons that we currently observe of sources at co-moving distances of 10, 20, 30 and 40 Gly have redshifts of 0.87, 2.63, 8.20 and 53.22 respectively.
  • The edge of the observable universe is receding from us with a recession velocity of more than 3 times the speed of light. $3.18c$, to be exact. In other words, we can observe sources that are moving away from us faster than the speed of light. Sources at co-moving distances of 10, 20, 30 and 40 Gly are receding from us at 0.69, 1.38, 2.06 and 2.75 times the speed of light, respectively.
  • Sources outside our particle horizon are moving away even faster. There is no a priori limit to the maximum recession velocity: it is proportional to the size of the total universe, which could be infinite.
  • The Hubble distance lies completely inside the event horizon. It will asymptotically approach the event horizon (as well as the curve of constant redshift 1) as $t$ goes to infinity. The current Hubble distance is 14.5 Gly (corresponding with $z=1.48$) , while the current distance to the event horizon is 16.7 Gly ($z=1.87$). Photons emitted today by sources that are located between these two distances will still reach us at some time in the future.
  • Although the difference between the Hubble distance and the event horizon today is rather small, this difference was much larger in the past. Consider for example the photons that we observe today, emitted by a source at a co-moving distance of 30 Gly. It emitted those photons at $t=0.62$ Gy, when the source was moving away from us at $3.5c$. The source continued its path along the vertical dotted line, while the photons moved on our past light cone. At $t=0.83, 1.64, 4.06$ Gy those photons passed regions that were moving away from us at $3c, 2c, c$ respectively. Along the way, those photons accumulated a total redshift of 53.22.

From all the above, it should be clear that the Hubble distance is not a horizon. I should stress again that all these calculations are only valid for the standard ΛCDM-model.

Apologies for the very lengthy post, but I hope it has clarified a few things.

  • 12
    $\begingroup$ +1 for the Davis and Lineweaver reference - I can't cite that paper enough around here. Also awesome entire rest of the post! And a belated welcome to Physics Stackexchange! $\endgroup$
    – user10851
    Commented May 17, 2013 at 23:46
  • 3
    $\begingroup$ I find this to be a vacuous philosophical babbling. A velocity is always a coordinate-dependent concept, even in special relativity (and in Newtonian physics). That doesn't mean that we shouldn't talk about it. We must talk about it because it is a crucial concept to describe all motion in Nature. For objects separated by large regions of a curved spacetime, it becomes ambiguous and important what we exactly mean by a velocity, but it is still true that there is a correspondence with a situation in special relativity which is the reason why we don't see things behind the cosmic horizon. $\endgroup$ Commented Jul 7, 2013 at 6:01
  • 17
    $\begingroup$ @Lubos Vacuous philosophical babbling??? All my calculations are correct and well known to cosmologists. I expected better from you. $\endgroup$
    – Pulsar
    Commented Jul 7, 2013 at 13:20
  • 3
    $\begingroup$ +1 What a beautiful explanation and also what a magnificent job with the plot: what a work of art, and so clear too. How long did that take to put together? I propose you think about putting it alongside the isometric embedding plot at en.wikipedia.org/wiki/… as it would really help clarify what the reader is looking at. Having the isometric embedding is pretty funky, but your plot can put a great deal more info on essentially the same plot, so they would go well together. $\endgroup$ Commented Dec 17, 2013 at 7:09
  • 1
    $\begingroup$ @WetSavannaAnimal Thank you very much, I really appreciate that. Yes, it took quite a bit of time :-) I made my own cosmology notes and I wrote a Python programme for all the calculations, and to create plots like this; all in all a few weeks work. I've made similar plots for proper distances, which you can see here. As for wiki, that would be a good idea, I'll think about it. Thanks again! $\endgroup$
    – Pulsar
    Commented Dec 17, 2013 at 15:13

Yes, the expansion of space itself is allowed to exceed the speed-of-light limit because the speed-of-light limit only applies to regions where special relativity – a description of the spacetime as a flat geometry – applies. In the context of cosmology, especially a very fast expansion, special relativity doesn't apply because the curvature of the spacetime is large and essential.

The expansion of space makes the relative speed between two places/galaxies scale like $v=Hd$ where $H$ is the Hubble constant and $d$ is the distance. When this $v$ exceeds $c$, it means that the two places/galaxies are "behind the horizons of one another" so they can't observer each other anytime soon. But they're still allowed to exist.

In quantum gravity i.e. string theory, there may exist limits on the acceleration of the expansion but the relevant maximum acceleration is extreme – Planckian – and doesn't invalidate any process we know, not even those in cosmic inflation.

  • 1
    $\begingroup$ @Motl Then one can assume that sending messages between such galaxies is impossible, since they are receding from each other faster than the speed of light? How then they still affect each other by gravity, which has the speed of light? $\endgroup$
    – Force
    Commented Apr 9, 2013 at 13:46
  • $\begingroup$ Dear Jim, yes, as I said, the fact that the mutual speed exceeds the speed of light - although it's not really well-defined - means that they can't see each other or otherwise communicate. Otherwise, they also can't send signals to each other gravitationally - gravitational signals propagate by the speed of light, too. However, both faraway galaxies still feel the gravitational field - but in general relativity, gravity is given by the local curvature of the spacetime which is there regardless of what the other distant galaxy is doing right now. $\endgroup$ Commented Apr 9, 2013 at 19:05
  • 3
    $\begingroup$ "the fact that the mutual speed exceeds the speed of light - although it's not really well-defined - means that they can't see each other or otherwise communicate" Yes, it's not uniquely defined, and that's the fundamental answer to the question. However, for the most common cosmological definition of the speed, the remainder of this sentence is false. See arxiv.org/abs/astro-ph/0310808 . $\endgroup$
    – user4552
    Commented May 8, 2013 at 12:22
  • $\begingroup$ Dear Ben, maybe I am not using the "most common definition of the speed" in a curved spacetime, but the statement of mine is surely correct for the "most natural definition of the speed" in a curved spacetime for any comparison with special relativity. $\endgroup$ Commented Jul 7, 2013 at 5:57
  • 4
    $\begingroup$ Still, photons can travel between galaxies that are receding from each other faster than light already at the time of emission; see the Davis & Lineweaver paper linked by @Pulsar below. By the way, impressingly, Davis wrote this paper as part of her Ph.D. $\endgroup$
    – Thriveth
    Commented Dec 13, 2013 at 22:15

Your question is based on a fundamental misconception. You say:

At the beginning, right after the Big Bang, the universe was the size of a coin

but it's more accurate to say "the observable universe was the size of a coin" i.e. the 13.7 billion light year bit that we can currently see was at one time the same radius as a coin. The universe may well be infinite in size, and if so it has always been infinite in size right back to the moment of the Big Bang.

There is no point in the observable universe that is moving away from us at faster than the speed of light, but assuming the universe is infinite, or at least much bigger than the bit we can see, everything farther away from us than the edge of the observable universe is moving away from us faster than the speed of light. As Luboš says this doesn't violate relativity since it's space that's expanding not the objects themselves moving, and there is no limit to the expansion rate of space. In fact if there was a period of inflation immediately after the Big Bang, during this period space expanded at a rate that makes the speed of light look positively glacial.

If you're interested in a bit more detail about how we model the expansion of the universe search this site for "FLRW metric", or Google for it.

  • 12
    $\begingroup$ "There is no point in the observable universe that is moving away from us at faster than the speed of light[...]" This is a common misconception. See Davis and Lineweaver arxiv.org/abs/astro-ph/0310808 . The answer by Pulsar is the only one that addresses the fundamental issue inherent the question, which is that GR doesn't have a well defined notion of the velocity of one object relative to some other distant object. (The Davis and Lineweaver uses one particular definition in the context of cosmology.) $\endgroup$
    – user4552
    Commented May 8, 2013 at 12:20

I'm going to go with "yes, but that's less interesting than you may think."

1. The laws of physics are local

Every law of physics that we know of only "sees" a tiny portion of the universe. The universe seems to consist of the same physical laws being applied identically and independently to every little part of itself.

If you look at any tiny part of an expanding universe, nothing untoward is going on. Everything is following the same laws as in any other situation, and nothing is exceeding the speed of light. When you stitch all of these pieces together, you get a global spacetime where the total volume of space seems to increase very quickly, but this "total volume of space" doesn't appear in any physical law, and in some sense you could think of it as a human invention.

2. Even globally, it's not clear that anything untoward is going on

The Milne model is the zero-density limit of the standard (FLRW) expanding cosmological model. It's a useful source of counterexamples to misconceptions about cosmology because it's actually just a portion of Minkowski space (the flat spacetime of special relativity) in different coordinates, so you can apply your special-relativistic intuition and calculational techniques to problems in cosmology, often getting results that contradict what might appear to be true in the FLRW coordinates.

In the Milne model, recessional velocities between objects can be arbitrarily high (exceeding $c$ or any particular multiple of $c$). This doesn't contradict special relativity because the definition of "recessional velocity" doesn't match the usual definition of "velocity" in special relativity. The recessional velocity is, in SR terms, the rapidity (times $c$).

In the Milne model, you and your friend (both at rest relative to the Hubble flow) can be 1 meter apart, wait for 1 second (measured by your respective watches), and at the end of that second be 10100 meters apart – or any other time interval and two distances you like, as long as the later distance is larger than the earlier one.

How can there possibly be "room" for this in Minkowski space? It's pretty easy to see what's going on. Since any inertial frame is as valid as any other, I'll pick one where you and your friend have equal and opposite velocities $\pm \mathbf v$. After a time $\tau$ has elapsed on your watches, your $t$ coordinates will have increased by $\gamma\tau$ and your x coordinate by $\pm\mathbf v\gamma\tau$. Since $\gamma\to\infty$ as $|\mathbf v|\to c$, these coordinate changes can be arbitrarily large, so there can be plenty of "room" at the end even if $\tau$ is small.

Another way of looking at this is that the triangle inequality doesn't work in spacetime. You might expect that if you and your friend start at the same point and you each travel in a straight line (inertial motion) for 1 second (elapsed proper time = length of worldline), that the distance between you should be at most 2 light seconds. In fact, though, the distance can be anything. If we classify that as "superluminal expansion of space" (and I think we should, since we are literally doing FLRW cosmology here), then superluminal expansion of space is allowed even in special relativity.

When you move from this special case to general FLRW cosmology, you lose the special-relativistic correspondence, but I don't think that makes the possibility of "superluminal" expansion any more surprising. On the contrary: if it can happen in special relativity, then of course it can happen in general relativity.

  • $\begingroup$ Very interesting. What about the approximation to an "empty" universe? $\endgroup$
    – Daniel
    Commented Nov 17, 2019 at 12:00
  • 1
    $\begingroup$ @Edouard There's no Newtonian gravity in this answer. The Milne cosmology isn't Newtonian. In Guth's toy Newtonian model, the collapse time doesn't depend on the radius of the universe, but it does depend on the density. That's true in FLRW cosmology too. You can cut off the FLRW geometry at some arbitrary radius (at least if $p=0$) and the time to the big crunch will be independent of the cutoff. It depends on local properties (density, Hubble parameter) and not on global properties (total mass, radius). $\endgroup$
    – benrg
    Commented Feb 2, 2021 at 22:07
  • $\begingroup$ @safesphere Yes, superluminal speeds are impossible in the sense that nothing can outrun a beam of light. Speeds larger than 299792458 m/s are possible if you define "speed" in certain ways, but nearby light also has a speed larger than 299792458 m/s by the same definition, so you still can't outrun it. I only talked about "superluminal expansion of space" at the end because other people do. I didn't mean to suggest it's really superluminal. We're in violent agreement, I think. $\endgroup$
    – benrg
    Commented Jun 30 at 6:11
  • $\begingroup$ @safesphere On your second point, I know you dislike my use of the term "Milne model" in various answers. The FLRW special case does differ at least philosophically from Milne's historical model. Wikipedia supports my usage, not that that means anything. I possibly should use a different name, but I'm not sure what name to use. $\endgroup$
    – benrg
    Commented Jun 30 at 6:13

I should add, in order to avoid inconsistencies in the text and in the derived formulas, that the expression for $D(z_{ob},t_{ob})$ as it written down here already IS a comoving distance, which forces you to set $a_{ob} = a(t_{0}) = 1$ and $t_{ob} = t_0$, $z_{ob} = 0$. This is a direct consequence of setting $ds^2 = 0$ in the FRLW line element, yielding the light-cone equation.

This distance is also the comoving distance given on the horizontal axis of the diagram. The text and the derived formulas should be adapted to these notions. For a correct treatment, refer to the Davis and Lineweaver papers. rhkail


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