Olbers' Paradox

Calling this a paradox is a slight courtesy. Nothing contradicts anything; a set of assumptions makes a prediction, and the prediction is false. The interest lies entirely in which assumption fails, and that question has a longer and more tangled history than the puzzle itself.

Which assumption is on trial. The argument needs all of the following: space is infinite (or at least very large), it is uniformly filled with stars, it is static, and it is infinitely old. Reject any one and the blaze goes away. Different centuries rejected different ones.

The expansion answer had a good reason to exist. In the steady-state cosmology of Hermann Bondi, Thomas Gold and Fred Hoyle, proposed in 1948, the universe genuinely is infinitely old and infinitely large. There is no finite age to appeal to. In that model the redshift really is the resolution — the only one available. Bondi's 1952 textbook Cosmology discussed the problem at length and fixed the name "Olbers' paradox" in the literature, and it presented the expansion answer because his cosmology required it.

Steady state lost. The finite-age answer was left standing, but the expansion answer had already propagated into a generation of textbooks and stayed there.

Harrison's correction. Edward Harrison argued from the 1960s onward, and at book length in Darkness at Night (1987), that even in a static universe of finite age the sky would be just as dark, and that the redshift changes the answer by a factor of order two against a required factor of $10^{13}$. Quantitative work in the 1980s by Paul Wesson and collaborators put numbers to this. The consensus since is Harrison's: finite age dominates, expansion is a minor correction, and absorption is no answer at all.

The escape nobody takes seriously any more. You can also break the argument by giving up uniformity. If stars cluster into galaxies, galaxies into clusters, clusters into superclusters, and so on without limit — a hierarchical or fractal universe, proposed by Carl Charlier in 1908 and revived by Benoît Mandelbrot in the 1970s — then the average density of matter falls to zero at large scales and the lines of sight escape. It is a real mathematical escape. It is also contradicted by observation: on scales above a few hundred million light years the universe looks homogeneous, not fractal.

Why it stays interesting. Almost every popular account of this puzzle gets the resolution wrong, usually by promoting the redshift or by keeping Olbers' own dust. That the correct answer was written down by a poet in 1848 and by Lord Kelvin in 1901, and still had to be argued for in the 1980s, says something uncomfortable about how explanations propagate.

1576. Thomas Digges published an English rendering of Copernicus with a diagram of stars scattered through unbounded space — one of the first infinite-universe cosmologies in print. He noticed the difficulty and waved it away: the distant stars, he said, are simply too faint to see. This is the wrong answer, but he was the first to feel the need for one.

1610. Johannes Kepler, replying to Galileo's Sidereus Nuncius in his Dissertatio cum Nuncio Sidereo, stated the argument properly for the first time. If stars filled an endless space the whole sky would shine; it does not; therefore the realm of stars is bounded and there is a dark wall around it. Kepler took the darkness as evidence, which is precisely the right instinct.

1721. Edmond Halley read two papers to the Royal Society on the infinity of the sphere of fixed stars. He tried to escape by arguing that remote stars subtend such small angles that their light becomes negligible. The argument does not work — as we saw, the shrinking is exactly cancelled by the growing number of stars — and Halley appears not to have noticed.

1744. Jean-Philippe Loys de Chéseaux, in an appendix to his study of the great comet of 1743, gave the first quantitative treatment. He computed how far one must look before the stars tile the sky, obtained an enormous number, and proposed that a thin absorbing fluid in space removes the excess light.

1823. Heinrich Olbers published Ueber die Durchsichtigkeit des Weltraums — "On the transparency of space" — in Bode's astronomical yearbook. He restated the problem clearly and proposed, apparently independently of de Chéseaux, the same absorbing medium. He is the only one of these men to have the puzzle named after him, and the answer he gave is wrong. Harrison's verdict is that Olbers' contribution was mostly to be the most-read statement of a problem others had posed better.

1848. Edgar Allan Poe, in Eureka: A Prose Poem, wrote the passage that should have ended it:

Were the succession of stars endless, then the background of the sky would present us a uniform luminosity, like that displayed by the Galaxy — since there could be absolutely no point, in all that background, at which would not exist a star. The only mode, therefore, in which, under such a state of affairs, we could comprehend the voids which our telescopes find in innumerable directions, would be by supposing the distance of the invisible background so immense that no ray from it has yet been able to reach us at all.

That is the finite-age resolution, correct in outline, written by a poet with no physics to support it, sixty years before anyone had a reason to believe the universe had a beginning. Poe was not doing science and should not be credited with a discovery. He should be credited with the sentence.

1901. William Thomson, Lord Kelvin, in an essay on ether and gravitational matter through infinite space, gave what is essentially the modern quantitative argument: stars have finite lifetimes and have been shining for a finite time, so only a limited shell of them can contribute. It was largely overlooked, and Harrison's rediscovery of it in the 1980s is one of the small pleasures of the literature.

1948–1952. The steady-state cosmology forces attention back to the problem; Bondi's Cosmology names it after Olbers and answers it with expansion.

1964–1987. Harrison develops and then publishes the finite-age account, recovers the Poe and Kelvin priority, and settles the question.

It is cosmology you can do with your own eyes. Before there were telescopes capable of measuring redshifts, before anyone had heard of a Big Bang, the plain fact that it gets dark at night ruled out the simplest imaginable universe: infinite, unchanging, eternal, evenly filled. That is a real empirical constraint derived from an observation every human being has made.

It made the finite age of the universe thinkable. For most of the history of Western cosmology an eternal universe was the default, and a universe with a beginning was a theological claim rather than a physical one. The dark sky is an argument, available since 1610, that the default cannot be right. It was not recognised as such until the twentieth century, which is part of its interest.

It introduced the idea of a horizon. The modern answer — we see a finite universe because light is slow and the universe is young — is the first appearance of the observational horizon that now runs through all of cosmology. Everything we can ever learn about the universe comes from inside a sphere defined by how long light has had to travel. Olbers' paradox is the oldest consequence of that fact.

It is a case study in a wrong answer surviving. The dust explanation is memorable, intuitive, given by the man the paradox is named after, and false. It outlived its own refutation by roughly a century and a half, and still appears in print.

The microwave background is Olbers' sky. The prediction was right: there is a uniform glow filling every direction. It comes from the hot plasma of the young universe rather than from stars, and expansion has stretched it from white-hot to 2.7 kelvin. Measuring the tiny variations in that glow — COBE in 1992, WMAP, Planck — is how the age, geometry and composition of the universe are now determined. The paradox's answer became its own most productive observation.

Counting all the starlight ever emitted. The extragalactic background light is the integrated output of every star that has ever shone, and it is exactly the quantity Olbers' argument is about. It is faint and very hard to measure directly because foreground zodiacal light swamps it. Astronomers now measure it indirectly: high-energy gamma rays from distant blazars are absorbed by collisions with background photons on the way here, so the attenuation of the gamma-ray spectrum encodes the background's intensity. The Fermi Gamma-ray Space Telescope collaboration used this in 2018 to reconstruct the history of star formation across cosmic time. The night sky is dark, and precisely how dark tells you when the stars turned on.

The shape of the argument keeps recurring. "If X were common, we would already see it; we do not; therefore something limits X" is now a standard move in astronomy — used on interstellar civilisations, on primordial black holes, on Population III stars, on cosmic strings. Olbers' paradox is the first time it worked, and also the first demonstration that the conclusion can turn out to be about the observer's horizon rather than about the thing being counted.

The divergence. Let stars have number density $n$, mean luminosity $L$, and sit in a static, infinitely old, Euclidean, uniformly filled universe. A shell of radius $r$ and thickness $dr$ contains $4\pi r^2 n,dr$ stars, each delivering a flux $L/4\pi r^2$ at the observer. The total flux received from within radius $R$ is

$$\Phi(R) = \int_0^R \left(4\pi r^2 n,dr\right)\frac{L}{4\pi r^2} = n L R .$$

The $r^2$ factors cancel exactly. $\Phi$ grows without bound, linearly in $R$.

The saturation. Stars are not points. Give each a cross-section $\sigma$. The probability that a line of sight reaches radius $r$ unobstructed is $e^{-r/\lambda}$ with mean free path

$$\lambda = \frac{1}{n\sigma}.$$

The sky's surface brightness then saturates rather than diverging:

$$\Sigma_{\text{sky}} = \Sigma_\star\left(1 - e^{-R/\lambda}\right),$$

where $\Sigma_\star$ is the surface brightness of a typical stellar surface. For $R \gg \lambda$ the sky is as bright as the Sun's disc, in every direction. For $R \ll \lambda$ it is fainter by the factor $R/\lambda$.

Putting numbers in. Using the observed density of stars in galaxies and the observed density of galaxies, Harrison estimates

$$\lambda \sim 10^{23}\ \text{light years}.$$

The horizon set by the finite age of the universe gives $R \sim 10^{10}$ light years. Hence

$$\frac{\Sigma_{\text{sky}}}{\Sigma_\star} \approx \frac{R}{\lambda} \sim 10^{-13}.$$

The observed extragalactic background is of that order. This single ratio is the resolution of the paradox, and note that no expansion appears anywhere in it.

The energy bound. There is a second, independent ceiling. Stars convert a fraction $\varepsilon \approx 0.007$ of rest mass into radiation, and only a small fraction $f$ of baryons have ever been processed through stars. The energy density of starlight is therefore bounded by

$$u_\star \lesssim f,\varepsilon,\rho_b c^2 \ll \rho_b c^2 ,$$

which is far below the $u = aT_\star^4$ required for a sky at stellar temperature. Even with unlimited time, the universe cannot afford Olbers' sky. The finite lifetime of stars and the finite age of the universe are two separate reasons, and either alone suffices.

What expansion actually contributes. For a source at redshift $z$, bolometric surface brightness is suppressed by $(1+z)^{-4}$ — one factor of $(1+z)$ for photon energy, one for arrival rate, and two from the relation between angular diameter and luminosity distance. Integrating this over the emission history of a realistic cosmology reduces the background by a factor of order unity — roughly two, not thirteen orders of magnitude. Setting the expansion rate to zero while holding the age fixed leaves the sky dark. Setting the age to infinity while holding the expansion leaves it bright. That asymmetry is the whole argument for finite age as the resolution.

Why absorption fails, formally. In a static universe of infinite age the radiation field reaches thermal equilibrium with the matter in it. Kirchhoff's law requires that a body's emissivity equal its absorptivity at every frequency, so at equilibrium any absorbing medium radiates exactly what it absorbs. The opacity redistributes the light; it cannot remove it. The absorbing medium ends at the temperature of the stellar surfaces and the sky is just as bright, now including the dust.