Observing the Multiverse (Guest Post)

It’s a big universe out there — maybe bigger that we think. A lot of people these days are contemplating the possibility that the wider world isn’t just more of the same; it could be that there are regions very different from ours, even with different low-energy laws of physics, outside our observable universe. It’s an old idea, which we now label the “multiverse,” even though we’re talking about regions of space connected to ours. A lot of other people are aghast that this is considered science. Personally I think science talks about unobservable things all the time, and this question is going to be resolved by people doing hard work to make sense of multiverse scenarios rather than by pronouncements about what is or is not science.

Matt Johnson We’re very happy to have a guest post from one of the people who is doing exactly that hard work — Matt Johnson, who guest-blogged for us before. He and his collaborators just come out two papers that examine the cosmic microwave background, looking for evidence of “bubble collisions.”

First Observational Tests of Eternal Inflation
Stephen M. Feeney (UCL), Matthew C. Johnson (Perimeter Institute), Daniel J. Mortlock (Imperial College London), Hiranya V. Peiris (UCL)

First Observational Tests of Eternal Inflation: Analysis Methods and WMAP 7-Year Results
Stephen M. Feeney (UCL), Matthew C. Johnson (Perimeter Institute), Daniel J. Mortlock (Imperial College London), Hiranya V. Peiris (UCL)

The hope is that these other “universes” might not be completely separate from our own — maybe we collided in the past. They’ve done a very careful job going through the data, with intriguing but inconclusive results. (See also Backreaction.)

Looking for this kind of signature in the CMB is certainly reminiscent of the concentric circles predicted by Gurzadyan and Penrose. But despite the similarities, it’s different in crucial ways — different theory, different phenomenon leading to the signal, different analysis, different conclusions. The road to sorting out this multiverse stuff is long and treacherous, but our brave cosmological explorers will eventually guide us through.

Here’s Matt.


Observing other universes: is this science fiction?

Perhaps not. Stephen Feeney, Daniel Mortlock, Hiranya Peiris and I recently performed an observational search for the signatures of colliding bubble universes in the cosmic microwave background. Before getting to our results, let me explain some of the back-story.

The idea that there might be other universes is taken quite seriously in high energy physics and cosmology these days. This is mainly due to the fact that the laws of physics, and the various “fundamental” constants appearing in them, could have been otherwise. More technically worded, there is no unique vacuum in theories of high energy physics that involve spontaneous symmetry breaking, extra dimensions, or supersymmetry. Having a bunch of vacua around is interesting, but to what extent are they actually realized in nature? Surprisingly, when a spacetime region undergoing inflation is metastable, there are cases when all of the vacua in a theory can be realized in different places and at different times. This phenomenon is known as eternal inflation. In an inflating universe, if a region is in a metastable vacuum, bubbles containing different vacua will form. These bubbles then expand, and eat into the original vacuum. However, if the space between bubbles is expanding fast enough, they never merge completely. There is always more volume to convert into different vacua through bubble formation, and the original vacuum never disappears: inflation becomes eternal. In the theory of eternal inflation, our entire observable universe resides inside one of these bubbles. Other bubbles will contain other universes. In this precise sense, many theories of high energy physics seem to predict the existence of other universes.

In the past four years, a few groups have tried to understand if it is possible to confront this radical picture of a “multiverse” with observation. The idea is to look for signatures of a collision between another bubble universe and our own. Even though the outside eternally inflating spacetime prevents all bubbles from merging, there will be many collisions between bubbles. How many we are even in principle able to see depends in detail on the underlying theory, and given the proliferation of theories, there is no concrete prediction.

Currently, the best information about the primordial universe comes from the cosmic microwave background (CMB). A collision will produce inhomogeneities in the early stages of cosmology inside our bubble, which are then imprinted as temperature and polarization fluctuations of the CMB. One can look for these fingerprints of a bubble collision in data from the WMAP or Planck satellites.

Most of the previous work has been to establish a proof of concept that observable bubble collisions can exist, and that there are theories which predict that we expect to see them; many of the details remain to be worked out. There are however a number of generic signatures of bubble collisions that we used to guide our search. Since a collision affects only a portion of our bubble interior, and because the colliding bubbles are nearly spherical, the signal is confined to a disc on the CMB sky (imagine two merging soap bubbles; the intersection is a ring). The effect of the collision inside the disc is very broad because it has been stretched by inflation. In addition, there might be a jump in the temperature at the boundary of the disc (although the magnitude and sharpness of such a jump has yet to be worked out in detail).

In a pair of papers (summary: arXiv:1012.1995 , details: arXiv:1012.3667) with Stephen Feeney, Daniel Mortlock, and Hiranya Peiris, we performed a search for these types of generic signatures in CMB data from the WMAP satellite. Our philosophy was to define a phenomenological model that encompasses the generic signatures of bubble collisions, and use the data to constrain the free parameters in the model. See the picture shown below, which is a simulated CMB sky containing a bubble collision, for an example of what a very clear signal might look like.

Simulated bubble

Predicted signal on the cosmic microwave background from a simulated collision with a bubble from another “universe.”

Simulated bubble collision

Zoomed in on the simulated bubble from above.

Cutting to the chase, we were first able to use simulated CMB data containing bubble collisions to rule out a range of parameter space as inconsistent with WMAP data. As it turned out, the existence of a temperature discontinuity at the boundary of the disc greatly increases our ability to make a detection. We did not find any circular temperature discontinuities in the WMAP data.

While we didn’t make any clear detections of bubble collisions, we did find four features in the WMAP data that are better explained by the bubble collision hypothesis than by the standard hypothesis of fluctuations in a nearly Gaussian field. We assess which of the two models better explain the data by evaluating the Bayesian evidence for each. The evidence correctly accounts for the fact that a more complex model (the bubble collisions, in this case) will generally fit the data better simply because it has more free parameters. This is the self-consistent statistical equivalent of applying Ockham’s Razor. In addition, using information from multiple frequencies measured by the WMAP satellite and a simulation of the WMAP experiment, we didn’t find any evidence that these features can be attributed to astrophysical foregrounds or experimental systematics.

One of the features we identified is the famous Cold Spot, which has been claimed as evidence for a number of theories including textures, voids, primordial inhomogeneities, and various other candidates. A nice aspect of our approach is that it can be used to compare these hypotheses, without making arbitrary choices about which features in the CMB need explaining (focusing on the Cold Spot is an a posteriori choice). We haven’t done this yet, but plan to soon.

While identifying the four features consistent with being bubble collisions was an exciting result, these features are on the edge of our sensitivity thresholds, and so should be considered only as a hint that there might be bubble collisions to find in future data. The good news is that we can do much more with data from the Planck satellite, which has better resolution and lower noise than the WMAP experiment. There is also much better polarization information, which provides a complementary signal of bubble collisions (found by Czech et. al. – arXiv:1006.0832). We’ll be gearing up to analyze this data, and hopefully there will be more to the story then.

  1. The multiverse is not an inevitable prediction of String theory. The idea that one day a dynamical vacuum selection principle will be found (maybe of cosmological origin) that would reveal why our vacuum (and our vacuum alone) was selected and realized among the others , has not been abandoned.

  2. Pingback: Tweets that mention Observing the Multiverse (Guest Post) | Cosmic Variance | Discover Magazine -- Topsy.com

  3. This fascinating analysis is reminiscent of a related search for which there was a flurry of activity (and popular articles and books) a number of years ago: analyzing the CMB for evidence of a non-flat topology to the observable universe – whether it might twist back on itself in some complex way and be shaped like a soccer ball with repeated patches, or some sort of funnel-like-thingy. I don’t recall if anything ever came of that search – anybody know? has it been substantially ruled out?

  4. Thanks for this post, it’s excellent—wonderful to hear first hand where this line of research is and where it is headed. But I have a couple questions:

    When you say “this is mainly due to the fact that the laws of physics, and the various “fundamental” constants appearing in them, could have been otherwise,” how do we know that?

    I understand our current theories give no deeper reason for the value of these constants, but how does that indicate they could have been different? Isn’t there a very real possibility that a deeper explanation of these constants exists, and that such an explanation leaves no room for variation, much like there is nothing we can do to alter the value of pi?

    I haven’t studied high energy physics, so I don’t really follow the more technical non-unique vacua explanation, (though I suppose it’s probably akin to the Dirac equation having solutions that suggested/predicted antimatter?), so maybe that explanation is much more compelling?

    In any case, thanks again!

  5. Until we find good evidence for these other multiverses, these ideas, while entertaining, should be considered ‘metaphysics’. Good luck with that.

  6. Before proceeding to the characterization of non-uniform CMB features, could you please first statistically eliminate the hypothesis that the CMB is uniform.

    Until the hypothesis that the CMB is uniform has been statistically demonstrated to be false, other inferences and hypotheses are moot.

    There are many statistical tests for non-randomness, especially in the cryptography and random number generation literature (e.g. information entropy measures), I would hope some of them would be applied to the CMB data to first test the hypothesis of uniformity; before proceeding to characterizing the non-uniformity.

  7. Okay I just realized I’m not completely clear on what I would have like to have seen in the arXiv articles. So specifically in the introduction I would have like to have seen a sentence something like:

    Previous tests for non-uniformity in the WMAP dataset statistically eliminate effect sizes greater than or equal to X at a significance of Y (insert citations), this leave open the possibility of detecting non-uniformities of characteristic angular size Z and temperature variation T or smaller, this paper analyzes …

  8. Is there any way to restate the statistical significance of these results in the way this is conventionally done in particle physics experiments? In other words, can your results be interpreted as a three-sigma “observation” of a bubble-collision from another universe, or as a five-sigma “discovery” of such a thing?

  9. Aaron (@6 & 7)

    I am not quite sure what you mean… when WMAP released its first dataset it showed variation at the level of 10^{-5}. The period from discovery (~1960) to mid-1980s there was no evidence for non-uniformity in the CMB as measurements were not good enough. BOOMERANG, MAXIMA and taco were the first experiments to see deviations from uniformity.

    The Planck or Wmap papers would be a good place to start looking. If you meant non-gaussianty rather than non-uniformity then we are still waiting for a verdict.

  10. Hi Matt,

    Thanks for taking the time to blog. I’m still wondering how plausible is it to expect a signal in the suitable parameter range to begin with? Is there some range of parameters that would seem natural? Best,


  11. @Cody: The thing is that pi and, say, epsilon-naught (aka the permittivity of free space – damn the lack of an epsilon key!) really are different. The value of pi jumps out of math and can’t be a different value – if the state legislature were to define pi as 22/7, math would stop working. Pi can no more take on different values than 1 can. On the other hand, the value for permittivity and many other fundamental constants are not something you can solve for – you have to measure them. And physics would still work perfectly fine if they had different values. Of course, the nature of the universe would change quite a bit, but you wouldn’t break physics if the values of these constants changed.

    So the question then becomes: why these particular values and not others? With most aspects of nature, everything not mandatory is forbidden: in other words, if a certain variation in a property is allowed, we expect to see the variation. If we don’t see the variation, we begin to suspect that such variation is not, in fact, allowed. So given that we have these fundamental “constants” that could really be parameters, we’re driven to find out whether there’s a) some constraining rule that we’re missing, or 2) look for somewhere where the values are different.

  12. I have a lovely theory (whose experimental support is as strong as that for superstrings) and it predicts that when god was creating the universe some of the angels were so dazzled by his awesome power that they collided, and guess what, those collisions left certain signatures in CMB (the precise type of signature depends on the type of angel).

    Now all I need to do is find those signatures in CMB, and everyone will have to accept my theory, right? RIGHT?

  13. Giotis – I’d love it if such a vacuum selection principle existed. However, at this point, I don’t think there is one. Working with what we know about string theory, the eternal inflation scenario can plausibly be realized. That is good enough motivation in my opinion to explore these ideas in the context of string theory (additionally, you don’t need string theory for eternal inflation, so this is an interesting phenomenon in its own right). However, one should certainly keep an open mind for what other possibilities might exist.

    Cody – In the context of string theory, the many “vacua” I am referring to arise from the many ways one can imagine hiding the extra dimensions predicted by string theory. This is typically accomplished by compactifying the extra dimensions. The shape and size of the compactified extra dimensions (along with other ingredients like branes, etc) determines the properties of the four-dimensional physics we know and love. So, for every shape and size, one obtains a different set of physical constants (like the planck scale, vacuum energy, etc), and perhaps a different set of interactions (perhaps there is another force in other vacua on top of the four we observe in our own universe).

    Aaron – The CMB is certainly anisotropic. One typically characterizes the statistics of these anisotropies in terms of the power spectrum (and higher point correlation functions). Tests have been applied to determine if there is any variation on the sky in these statistics, yielding some evidence for statistical anisotropy. However, if one is looking for a local fluctuation in the temperature of the CMB of a very specific form (as we were), these types of statistics are typically not the most useful thing to look at. Indeed, many things could be buried in the data which would, say, yield a negligible “bump” in the power spectrum compared to experimental errors and cosmic variance (ie the small sample size of large-scale modes). This perhaps gets at the sentence you wanted to see in our paper: Before doing our analysis, we can rule out features in the data with an amplitude much larger than the characteristic amplitude of features set by the power spectrum (around 10^{-4} kelvin or so), or other measured statistical properties of the CMB.

    Peter – We comment briefly on how one can obtain a “sigma” from the Bayesian evidence ratio in the longer paper (see the discussion near Eq. 28). There are various assumptions made in doing so, but very roughly, you can interpret our results as near 3-sigma.

    Bee – The probability that we 1) have a bubble collision in our past and 2) it yields a signal strength consistent with current data, is highly uncertain. It could very well be negligible. For detailed discussion of these issues, you can check out the review paper I wrote with Anthony Aguirre (arXiv:0908.4105). What previous work has shown is that the criteria for observing bubble collisions can be satisfied in some cases. Without a more detailed picture of, say, the string theory landscape, it is difficult to ascertain the theoretical prior on what to expect. Nevertheless, we have the data, some idea of what the signature of bubble collisions should look like, and some theories where we expect to see them, so our philosophy was: they might be there, so why not go look?

  14. Multiverse ‘physics’ is NOT science. That is an absolute fact and an absolute waste of some great minds.

  15. Thanks Sean, but what I was trying to say was that I don’t see any good reason to expect variation in our fundamental constants to be allowed. I understand π can’t be different, nor 1, and I am uncomfortable assuming ε_0 can be different (though I know current theory leaves ~25 constants to empirical measurement—I just think it more likely there is a deeper mathematical origin we have yet to find; further constraints as you mentioned).

    Meaning, if I had to wager whether the constants must be what they are, or whether they can vary in a grander multiverse, I’d bet against the latter. But I do see how, at the moment, variation in the fundamental constants is an open question and that my position (searching for a deeper cause for the constants, in say number theory) is stagnant, and also that there are new directions to search for variation (as Matt has described above).

    So although I would more naturally side with your option “a”, I can see how option “2” is currently a more active and productive area of research.

    Matt, thanks for expanding on the vacuum bit.

  16. Hi Matt,
    Thanks for the reply. Sure, I understand the let’s just go and look approach. I was just wondering if the parameter range that can be tackled with today’s data is one that’s plausible to expect for such an effect and couldn’t find much on it. See, I’m thinking if it was extremely unlikely that eternal inflation would produce such a signal to begin with, then it’s not much of a constraint, is it? Will check out the paper. Merry Christmas,


  17. Non-physicist here: If two such bubbles can collide, does that then mean that these bubbles have a bounding edge (as in the soap bubble analogy)? Since our universe has no edge in the sense of a brick wall, does it have one somewhere else? What type of edge do these bubbles have? What do these bubbles exist within?

  18. What if this or other micro-arrangement of past produced patterns is not collision or ejection, perhaps a continuation, or initiation of an unobservable inflationary phenomenon outside of our Universe, yet still in some minor ways connected or…just a pattern.

    Keep looking up and Happy New Year.

  19. I have never been one to believe in the existence of “only one”. If one can be created, more must have been created, just like our solar system is not alone within our galaxy, within our universe.

    This is pure speculation, but the mystery of the missing antimatter from the Big Bang may be explained by forming a separate antimatter universe in conjunction with ours. During the period of inflation, the universe expanded much faster than the speed of light. If the two universes; one matter, the other antimatter; were driven apart during the inflationary period, we would not be able to observe the second universe because the light from it would not have reached us. I wonder if there is such a thing as anti-light. There may well be several universes we cannot see or may have seen, but not recognized.

    Food for thought – Bill

  20. Sean, Matt and others:
    what do you think of http://arxiv.org/abs/1007.0587
    Although this has got a lot of press , have not seen any reactions from physicists.
    (also don’t think torsion has been explored in cosmological context much)

  21. I’m not sure why that paper got so much attention. Part of it is speculative but fairly straightforward, then the part about new universes is just hand-waving.