What is the universe expanding into?

Edwin Hubble reported in 1929 that galaxies recede at speeds roughly proportional to their distance. A uniform expansion produces exactly that relation, and it produces it identically from every point, so no observer is at a centre and there does not have to be one. An explosion would have one.

What Penzias and Wilson reported in 1965 was an excess: 3.5 kelvin of antenna noise they could not account for, isotropic, unpolarised and free of seasonal variation. They offered no explanation of their own, and pointed instead at the letter printed alongside theirs for a possible one. That letter proposed black-body radiation left over from a hot, dense early universe and cooled by its expansion. In the standard model the radiation dates from about 380,000 years after the Big Bang, when the plasma recombined and the universe went transparent.

The pattern of its tiny variations is now measured to extraordinary precision, and it encodes the composition, geometry and age of the universe. The age comes out at 13.80 billion years, give or take 23 million. The geometry is flat to within measurement error once the microwave background is combined with galaxy surveys, though Planck’s temperature and polarisation spectra on their own lean slightly closed, and adding its lensing measurement pulls them back toward flat.

In general relativity the distance between observers at rest with respect to the matter filling the universe grows with time: they separate without moving through anything. The universe does not expand into previously existing empty space. It consists of expanding space.

How hard that picture can be leaned on is argued about among cosmologists, because expanding space is a framework and not a mechanism, and reading it as a force is where the confusion usually starts. The geometry underneath the argument is not in dispute.

Two methods for measuring the current expansion rate disagree. Planck, working forward from the early universe under the standard model, gives 67.36 plus or minus 0.54 kilometres per second per megaparsec. Cepheids and supernovae in the nearby universe give 73.04 plus or minus 1.04. That team puts the gap at five sigma and reports no sign that measurement error or any of its analysis variations explains it.

Not every local measurement agrees with it. The Chicago-Carnegie Hubble Program, calibrating the supernovae with red giant stars instead of Cepheids, gets 70.39, with a larger uncertainty, and reports no need for new physics. The Cepheid team answers that the difference comes from which supernovae each sample happens to include.

A flat universe can be infinite in extent or finite with a non-trivial topology, like a shape that wraps around on itself. Planck searched the microwave background for the matched circles such a shape would print on the sky and found none. For a cubic torus, that puts the largest sphere fitting inside the cube the space repeats at a radius of more than 97 per cent of the distance to the surface of last scattering. A lower bound on the size, not an answer, and only for the shapes tested: later work found that in all but the simplest shapes the shortest loop around the universe can miss us and be shorter than the diameter of that surface by a factor of 2 to at least 6.

We can only see out to the distance light has had time to travel, and in practice only as far as the surface where the universe went transparent. Beyond that, the standard model assumes more of the same, and the assumption cannot be checked.

Receding faster than light is not what puts a region out of reach, which is the point at which this usually goes wrong. Everything beyond redshift 1.46 recedes superluminally, and such objects are observed routinely, because light that starts out in a superluminal region can end up in a slower one and reach us: one survey’s latest catalogue alone holds over 1.2 million quasars beyond redshift 1.77. The limit that is permanent is the event horizon, a different boundary and a nearer one: nothing that happens beyond it now will ever be seen from here, though galaxies beyond it are still visible by light they sent long ago.

The long-term fate depends on the behaviour of dark energy, which is not understood. If it is a true constant, expansion continues and the observable universe empties out. If it is not a constant, other outcomes remain possible: bound structures torn apart if its density grows, or a recollapse if it decays below zero.

Which of those is live has moved recently. The Dark Energy Spectroscopic Instrument finds its baryon acoustic measurements fit better with an equation of state that evolves than with a constant: preferred at 3.1 sigma combined with the microwave background, and between 2.8 and 4.2 sigma with supernovae added, depending on which supernova sample is used. The collaboration states the alternative itself, which is an unknown systematic error in one or more of the data sets.

Corrections to the supernova samples since have narrowed that range: a 2026 reanalysis puts all three at 3.2 to 3.4 sigma, and the team behind the sample that gave 4.2 now reports 3.2, which it calls a weak preference. George Efstathiou has argued that a small offset between supernova samples could bring the most discrepant one into line with a constant.

Our description runs backwards reliably to about a second, where nucleosynthesis begins, and then it stops. The Particle Data Group puts the boundary there and says why: nothing from the epochs before it, the quark-hadron transition or electroweak symmetry breaking, has left a relic anyone has identified, so cosmology cannot be traced in detail any earlier. “The Big Bang” names the earliest state we can describe, not an explanation of how it came to be.

A brief period of extremely rapid expansion in the first instants would explain the universe’s uniformity and flatness, and the statistical pattern of fluctuations in the microwave background matches its predictions well. Planck finds the fluctuations consistent with slow-roll single-field inflation and the universe flat to high accuracy.

Many versions of inflation imply that it never stops everywhere, producing regions beyond our horizon with potentially different properties. Guth, one of the physicists who first proposed inflation, calls those regions unobservable in the same paper that argues for them. The general framework has strong observational support. The specific models, and their multiverse implications, do not.

Asking what the universe expands into may be like asking what is north of the North Pole: a grammatically valid question built on a spatial intuition that does not apply. Expanding space is the staple of almost all cosmological teaching, and this is the answer that comes with it. It is worth noticing that it is an answer about our concepts, not about the universe.