This article was first posted in two parts on Scientia Salon.
The multiverse concept is often derided as “unscientific” and an example of physicists indulging in metaphysical speculation of the sort they would usually deplore. For example commenters at Scientia Salon have said that the multiverse is “by definition not verifiable and thus outside the bounds of empirical science”, and that “advocates of multiverses seem to be in need of serious philosophical help”. 
Critics thus claim that the multiverse amounts to a leap of faith akin to a religious belief. Indeed, the religious often accuse atheistic scientists of inventing the multiverse purely to rebut the “fine-tuning” argument that they say points to a creator god (though the fine-tuning argument is readily refuted in several other ways, and anyhow physicists really don’t care enough about theology these days to let that worry them; further, the concepts leading to a multiverse were developed well before theologians started taking note of the issue).
The purpose of this article is to argue that the multiverse is an entirely scientific hypothesis, arrived at for good scientific reasons and arising out of testable and tested cosmological models. To be clear, I am not asserting that the multiverse has been proven true, even on the balance of probability, but I am asserting that it is a serious scientific concept that will eventually be accepted or rejected on scientific grounds.
Several different concepts could be labelled a “multiverse”, but I am advocating one particular multiverse concept, that arising from what cosmologists call the “eternal inflation” version of Big Bang cosmology.  I’ll outline why cosmologists have arrived at this model, which is now a mainstream account of the origin of our universe, and which leads naturally to a multiverse.
The “inflationary” model of the Big Bang was developed to explain observations of our universe and predictions from this model have since been verified, putting it on a sound footing. If a scientific theory predicts consequences A, B, C and D, and if we then verify that A, B and C are indeed the case, thus giving us confidence in the theory, then we have sound reasons for accepting D even if D cannot be directly verified. Indeed, we would be obliged to accept D unless we can construct another equally good explanation of A, B and C that does not entail D (see here for a fuller account of this argument).
The Big Bang model
As long ago as Isaac Newton people realised that a static universe doesn’t actually work. Stars and galaxies would fall together under gravity and thus they can no more be static than you could float a brick in mid air. But scientists ignored this — even Einstein rigged his equations to get a static universe — until Edwin Hubble produced empirical evidence that the galaxies are moving away from us. Observations tell us that our universe is expanding, and expanding in a uniform fashion as though space itself is expanding, carrying the galaxies along with it.
If one runs a uniform expansion backwards in time — say repeatedly halving all scale lengths and separations — then eventually one arrives at a state with everything in the same place. However, we expect our understanding of physics to have broken down before any such “singularity”, and in particular we expect “quantum gravity” effects to dominate at a scale length of 10–35 metres, called the “Planck length”. We do not have a working theory of quantum gravity, but we know enough about quantum mechanics to suppose that our universe originated with a scale-length of 10–35 metres as a quantum fluctuation in a pre-existing “state”.
Understanding this process, and knowing about the pre-existing state in which the quantum fluctuation might have occurred, are not needed for the multiverse model that I expound below. However, where one quantum fluctuation can occur so can another, and thus it is natural to suppose that there might be many other such universes. In particular, we have no strong reason to suppose that the quantum fluctuation that originated our universe was the origin of all things or of time itself — though equally we lack arguments against those possibilities, and if you want to argue for them then go ahead. 
The characteristic length of 10–35 metres leads to a characteristic time, being the time taken to cross that length at a speed limited by the speed of light, c, the fastest speed at which information can be transferred from one region to another. The value of c of 108 metres/second then gives a “Planck time” of 10–43 seconds. This leads to the concept of an “observable horizon”, being the furthest distance (= ct) from which light/information can have travelled to us given an age, t, of the universe.
Thus we conceive of our universe, having originated as a quantum fluctuation, obeying its natural tendency for length scales to expand (that being the solution of Einstein’s equations of General Relativity), and with the observable horizon also expanding as the universe ages.
Conventional cosmological models presume this universe to be spatially infinite in all directions (or, at least, so much larger than the observable horizon that we don’t have to worry about the observational consequences of any “edge”). Is there an alternative? One could argue that our original quantum fluctuation would have had some finite extent, and thus that our current universe would now have a much larger but still finite extent, and that somewhere beyond the observable horizon is a boundary to another “universe” originating in a separate quantum fluctuation, or to the “pre-universe” out of which our quantum fluctuation arose.
Alternatively, one could have a “wrapped-around universe”, where space has a finite extent but one can travel forever in any direction, like ants walking on the surface of a sphere. Observationally, the “curvature” of space in our universe is known to be very small (our universe is “flat” to 1% accuracy), and thus the scale of the wrap-around would have to be much bigger than our observable horizon. Still, if you wanted to argue that the original quantum-gravity fluctuation was self-contained, rather than being in a pre-existing state, then one could argue for a finite, wrapped-around universe along these lines. 
The multiverse model is often criticised as “unscientific” for invoking universes that can never be seen and thus making claims that can never be verified. But this applies just as much to all cosmological models, which are usually presumed to extend to infinity. All of them are thus postulating that the universe stretches well beyond the observable horizon, from where (owing to the finite speed of light) we can never obtain information to verify any hypotheses. This feature does not make the model unscientific. If we are simply using principles of parsimony to postulate more-of-the-same, beyond where we humans can personally see it, then we’re being entirely scientific.
Is the above account supported by evidence? Yes, very much so. The expansion of space is observed in the red-shift of light from distant galaxies (caused by the expansion of photons’ wavelengths as they travel through expanding space). In addition the Big Bang model predicts exactly what primordial elements were created in the expanding fireball of the first 1000 seconds. The abundances of Hydrogen and Helium-4, and of traces of Deuterium, Helium-3, and Lithium-7 are predicted to high accuracy — and when observers measured the abundances to test the Big Bang predictions they found an excellent match.
Further, the Big Bang model predicts the existence of “cosmic microwave background” (CMB) radiation, left over from when the ionised universe cooled enough to allow neutral atoms, 380,000 years after the Big Bang. After this prediction had been made the radiation was then found, exactly as predicted, stretched out into the microwave band by the subsequent expansion of space (and leading to the first of three Nobel Prizes so far awarded for work on Big Bang cosmology ).
The study of the CMB, a snapshot of how the universe was soon after the Big Bang, is now a major part of cosmology. Indeed, cosmology is on a sound empirical footing precisely because we can directly see what the early universe was like; since that time the universe has been transparent and thus we can observe photons that last interacted with matter in the early universe.
Modelling the tiny temperature fluctuations in the nearly-smooth CMB produces strong constraints on cosmological models. As measurements have improved, from the COBE satellite to NASA’s WMAP and recently ESA’s Planck satellite, at each step better data could have produced results incompatible with the Big Bang models, but at each step the concordance between theory and observation has got better and more remarkable.
The comparison of cosmological models with high-quality and detailed observations of the early universe has led to the “inflationary” version of the Big Bang. This hypothesises that, very early in the first second after the initial quantum fluctuation, only about 10–35 seconds later, our universe started an episode of exceptionally rapid exponential expansion, growing by a huge factor of about 1030 in only a tiny fraction of a second.
This “inflation” was proposed in order to explain several otherwise puzzling features of the universe, including (1) the similarity of the universe in opposite directions in the sky, (2) the exceptional smoothness of the CMB, (3) the fact that the universe appears to have a very close balance between the total amount of matter and energy and the expansion rate, thus giving space a near-zero “curvature” on the largest scales, and (4) the absence of heavy particles such as magnetic monopoles that would otherwise be expected to be seen. Thus inflation is motivated by strong empirical evidence. 
Further, after inflation had been proposed, it was used to predict the spectrum of fluctuations expected in the CMB. These temperature fluctuations originate as quantum fluctuations in the inflating field, later to be frozen into the CMB, and the inflationary model gives specific predictions of their form and power spectrum.
These predictions have now been compared to the better-and-better data from the WMAP and Planck satellites, and several ground-based experiments, and the result is again an exceptional agreement, giving strong support for inflation.
Further still, the inflationary model makes specific predictions about gravitational waves generated by the rapid inflationary episode. Recently the BICEP2 experiment has reported detecting these gravitational waves, imprinted on the CMB just as predicted, which would be an additional strong confirmation of inflation. These results are new and need confirming, but at a minimum they show that the inflationary model can be directly tested.
Thus the inflationary Big Bang model is robust, mainstream cosmology. The mechanism driving inflation is, however, less understood. Modern physics explains all known interactions in terms of four forces (gravity, electromagnetism, and the strong and weak nuclear forces), and at the exceptionally high temperatures in the very early Big Bang all four forces are expected to have existed in a “symmetric” state where they all acted similarly. As the universe expanded and cooled the symmetry was broken and the four forces developed the different characteristics that we see today. This process would have been analogous to the “phase transition” from a liquid (hot) state to a crystalline (cold) state.
Such a phase transition has a “latent heat of crystallisation” given out during the transition. It is, though, possible for material to get stuck in a “supercooled” state where it hasn’t yet made the transition, and thus has extra energy than expected for its temperature. It is this energy — thought to be associated with the phase transition that breaks up of the strong and electroweak forces — that is thought to drive the ultra-rapid expansion of the inflationary era.
The tricky bit of inflationary models is then getting the universe to drop out of the “supercooled” inflationary state (rather than being stuck in that state for ever), and thus give its energy into the hot Big Bang that then produces our universe. In order to get such models to work theorists have developed a scenario in which a quantum fluctuation can cause a limited region to drop out of the inflationary state, forming an expanding bubble of normal-state universe.
Thus, overall, we have our universe originating as a quantum fluctuation in the quantum-gravity era, at a scale of 10–43 seconds, leading to the exponentially expanding inflationary-state, followed by quantum fluctuations within the inflationary state, at a scale of 10–35 seconds, leading to bubbles of normal-state universe.
However, in such a scenario, the inflationary-state stuff surrounding the bubble will be expanding vastly faster than the normal-state bubble, and thus the size of the inflationary-state regions continue to increase, even as bubbles continually drop out into the normal state.
The result is called “eternal inflation”, a “Swiss cheese” mixture in which bubbles of normal universe are continually forming out of a surrounding and exponentially expanding inflationary state. One of those bubbles would be our universe, and that bubble would now have expanded to vastly larger than our observable horizon. Thus, the only things we can now see are in our bubble, our normal-state universe.
This is a multiverse scenario. It says that somewhere beyond our observable horizon there is a transition, beyond which is supercooled, inflationary-state stuff. And in that rapidly expanding vastness are other bubbles, other universes, like ours, but now separated from us by unfathomable distances.
If you don’t like the idea of a multiverse extending vastly beyond our observable horizon, or consider it to be unscientific, then realise that conventional cosmological models extend to infinity in much the same way. The stuff beyond our observable horizon is real, it is just a long distance away. There is no reason to declare such stuff not “real” just because of the finite value for the speed of light, which means that we humans can never receive information from those regions (especially since the location of that observable horizon depends entirely on where you are looking from, and it also continually recedes as you look at it).
The only sensible alternative to this multiverse idea is that our universe extends vastly beyond our observable horizon (to infinity?) and is normal-state all the way. Is that really preferable? If you do prefer normal-state all the way, then you have a problem in constructing an inflationary model that correctly makes the transition from the inflationary state to a normal state everywhere, especially given that any “transition front” would have a speed limited by c. If you know how to do that or if you know of a way of doing away with inflation altogether, and yet still explaining all the observations that motivated inflation, then go ahead and publish it (and produce an inflation-free model of the CMB power-spectrum that fits just as well as the one in the figure above).
As it is, we have strong observational and theoretical arguments that lead us to an eternal-inflation model of the Big Bang, and that eternal-inflation produces a multiverse. It is currently rather difficult to produce a model of the Big Bang that works, and that explains the observations, and which does not automatically produce a multiverse.
Admittedly we cannot observe those other universes, the other normal-state bubbles continually forming in the inflationary-state multiverse, but good, sensible and scientific reasons lead us to conclude that they likely exist. And, as stated early on, it is not necessary to empirically validate every prediction of a theory in order to have good confidence in that theory. To accept a theory and its implications all one needs is to have validated some of the predictions of theory, and to have established that overall the theory does a better job than any alternative that we know of.
As a comparison, no-one disputes the validity and scientific status of laws of gravity, that do an excellent job of predicting the times of future solar eclipses, just because we cannot verify those eclipse timings indefinitely into the future; and similarly it is not grounds to reject a model as unscientific just because we cannot verify its predictions indefinitely into the far distance.
Are the “physical constants” constant?
Now let’s ask a further question. Are those other bubble-universes in the multiverse just like ours? In what ways might they be different? How much scope is there for the bubbles to differ?
Above I used the analogy of a bubble dropping out of the inflationary state being akin to a liquid freezing. Consider a snowflake freezing in a high-up cloud. The freezing of each snowflake complies with the same underlying laws of physics, and yet each snowflake is different, with a related but distinct pattern. This tells us that some aspects of what we see are local “accidents”, variations allowed by the underlying laws but contingent on local circumstance.
The transition from inflationary state to normal state is thought to be due to the fundamental forces changing from a “symmetric” state, where they acted similarly, to a broken-symmetry state where they have different strengths. Further, physical “constants” such as the masses of particles and the values of electromagnetic charges are much the same thing as the strengths of the forces, since essentially they are all telling us how particles interact with each other.
Thus we can ask, are the values of the masses and charges of particles and the strengths of forces dictated by the fundamental physical laws, or are they local accidents, dependent on the local contingency of symmetry breaking early on in the Big Bang? We don’t know the answer to that, but if it is the latter then we would expect each bubble universe to be different in the same way that each snowflake is different, and thus to have different physical constants.
Note that — counter-intuitively to some who have not thought about it — the latter suggestion is more parsimonious. The correct interpretation of Occam’s razor is in terms of the information content needed to specify a model.  If you have to explicitly specify the couple-of-dozen fundamental constants of the standard model of particle physics, that takes a lot of information. It takes less information to say that values for the constants are strewn around at random. Afterall, no-one claims that every snowflake having a different pattern is deprecated under Occam’s razor, since everyone accepts that the individual patterns are accidents, variations allowed by deeper-level rules.
As for falsifiability, it would be unscientific to add information to a model that had no observational motivation or consequences, and thus was unfalsifiable. However, if we are simply extrapolating from an observationally motivated model, or even reducing the information content of our model, while ensuring that the remaining information is observationally motivated, then that is scientific, even if not all of the implications of the model can be tested.
Considering this idea one quickly arrives at the obvious point that we human observers could only find ourselves in a bubble that had parameters suitable to have produced us, and it may be that the vast majority of such bubbles would be too alien to support or produce us.
This is an entirely normal way of thinking. No-one nowadays supposes that there is some mechanism that ensures that an Earth-size planet gets placed in an orbit at the right distance from its star to allow it to have liquid water; we now know that extra-solar planets have a huge variety of orbits. We, of course, find ourselves on a planet suitable for us, but likely there are vast numbers of similar but uninhabited planets where the conditions are not right for life.
Similarly, no-one since Darwin would argue that we find polar bears in the Arctic and camels in deserts because that was carefully and deliberately arranged; rather, we understand that the local fauna is the contingent product of the local environment — a statement that applies just as much to the multiverse.
If you still find that line of reasoning unpalatable, realise that Steven Weinberg used it to predict that the value of the “dark energy” parameter in our universe would be small but non-zero. This prediction was made a decade before the observational detection of dark energy, at a time when most cosmologists assumed there was no dark energy. Then it was found, with a value in line with Weinberg’s prediction, but which is vastly smaller than given by attempts at calculating it from fundamental physics.  Weinberg’s verified prediction is currently the best explanation we have of the amount of dark energy in our universe.
At this point, with an infinite extent of expanding inflationary-state stuff, dotted with island-universe bubbles, with each universe having different values for the physical constants, we have a full-blown multiverse of the sort to give critics a fit of the vapours. Yet, everything above is solid scientific reasoning, motivated by and supported by observational evidence, and with already-demonstrated predictive power. That does not mean that it is true or proven, but it is a fully scientific concept and, regardless of the critics, the scenario is becoming increasingly accepted and mainstream among physicists.
A last remark: if the above is correct then in all likelihood we are near the middle of such a bubble and not near its edge. Yet there is some possibility that we are near an edge, and if we are then there would be nothing to stop us observing it if we looked far enough into the distant universe. In principle, the scenario is directly verifiable. 
 See comments to this Scientia Salon article.
 Thus I am ignoring here the issue of a multi-dimensional multiverse possibly arising out of M-theory.
 See Lawrence Krauss’s book A Universe from Nothing for an example of this argument.
 When dealing with general relativity and quantum gravity the notions of time and space depend on the observer, so in discussing whether the universe has a finite extent we should specify where we’re observing from. A black hole could look finite from the outside but infinite from the inside.
 These being the 1978 prize to Penzias and Wilson for the discovery of the CMB, the 2006 prize to Mather and Smoot for the discovery of structure in the CMB, and the 2011 prize to Perlmutter, Schmidt and Reiss for the discovery of “dark energy” through the acceleration of distant supernovae. If the BICEP2 result holds up there is a likelihood of a fourth prize for the idea of inflation.
 See this article of mine for a defence of Occam’s razor as a scientific concept.
 Naive calculations of the amount of dark energy expected give values a factor of about 10120 too big, perhaps the biggest error in the history of physics! Why is it that the actual value is 10120 times smaller than expected?
 And of course cosmologists are already looking for possible effects of “colliding bubbles”, which might be visible in the CMB. See this post for an example.