The Cosmic Microwave Background: the universe's baby photo

Here is a photograph of the universe when it was 380,000 years old, and it fits in a sentence: a uniform glow at 2.725 kelvin, the same in every direction, dappled with hot and cold patches that differ by one part in a hundred thousand. That is the whole image. It sounds like a near-blank sky, almost nothing happening. Yet those faint patches, a few millionths of a degree across the mean, are not noise and not random. They are sound. Pressure waves that were ringing through the infant universe got frozen in place the instant the cosmos turned transparent, and we are still looking at the snapshot.

The claim worth resisting is that "the oldest light" should be some exotic, scrambled, hard-to-read signal. It is the opposite. The cosmic microwave background is the cleanest, most boringly thermal spectrum nature has ever handed us, a blackbody so perfect that for thirty years no experiment could find a flaw in it. The information isn't in the spectrum. It's in the one-part-in-100,000 deviations from perfect smoothness, and decoding them turns a featureless glow into the most detailed measurement we have of what the universe is made of.

A glow from every direction at once

In 1964 Arno Penzias and Robert Wilson were trying to do radio astronomy with a horn antenna at Bell Labs, and they had a problem: a faint hiss in the receiver that would not go away. It came from every direction in the sky, day and night, every season, the same. They checked everything. They cooled the amplifier, they hunted for loose connections, they famously evicted a pair of pigeons nesting in the horn and scrubbed out the droppings. The hiss stayed. It corresponded to an excess antenna temperature of about 3.5 kelvin that had no business being there.

Thirty miles away at Princeton, a group led by Robert Dicke was building an antenna specifically to look for exactly this: the leftover heat of a hot early universe, predicted decades earlier and then more or less forgotten. When Penzias phoned to describe the nuisance, Dicke hung up and said to his team, "Boys, we've been scooped." The hiss was the universe. Penzias and Wilson had measured, by accident, the afterglow of the Big Bang, and it won them the 1978 Nobel Prize.

What they had found was light with a very particular character. Not starlight, not the glow of any object, but radiation with a thermal spectrum: the exact shape a perfect absorber and emitter produces when it sits at one temperature. That shape has a name and a formula.

$$I_\nu(\nu, T) = \frac{8\pi h \nu^3}{c^3}\,\frac{1}{e^{h\nu/k_B T} - 1}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

Here $\nu$ is frequency, $T$ the temperature, $h$ the Planck constant, $k_B$ the Boltzmann constant, and $c$ the speed of light. The curve rises, peaks, and falls, and its whole shape is fixed by a single number, $T$. Plug in $T = 2.725$ K and you get a peak near 160 GHz, a wavelength of about a millimetre, squarely in the microwave band. Wien's displacement law is the shorthand for where that peak sits.

$$\lambda_\text{peak}\, T = 2.898 \times 10^{-3}\ \text{m·K}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

In 1990 the COBE satellite's FIRAS instrument measured this spectrum across many frequencies at once and produced what is still one of the most famous plots in physics: the data points fall on the theoretical blackbody curve so exactly that the error bars are smaller than the width of the line. John Mather presented it to a room of astronomers and got a standing ovation, which is not a thing that normally happens to a spectrum. The deviations from a perfect Planck curve are under 0.005%. Nothing else in nature comes this close.

The day the universe turned transparent

Why was there a hot plasma, and why did it stop being one? Wind the clock back. The universe is expanding, so it was once smaller, denser, and hotter. Run far enough back and you reach a state where everything is a soup of protons, electrons, and photons at thousands of degrees. At that temperature electrons cannot stay bound to protons: every time one tries to settle into a hydrogen atom, a photon comes along and knocks it loose again. So the electrons roam free, and free electrons are extremely good at scattering light. A photon could not travel any meaningful distance before bouncing off an electron. The universe was a glowing fog.

Then expansion cooled it. Around 380,000 years after the Big Bang the temperature fell to roughly 3,000 K, and below that threshold the photons no longer had the energy to keep electrons stripped from protons. Electrons and protons paired up into neutral hydrogen, a transition cosmologists call recombination (a slightly odd name, since they had never been combined before). Suddenly there were almost no free electrons left to scatter light. The fog cleared. Photons that had been bouncing around for thousands of years streamed off in straight lines, and most of them have been travelling, untouched, ever since.

There is one more piece of bookkeeping. That fog was glowing at 3,000 K, the colour of a dim red star, not a chilly 2.725 K. So why is the CMB cold? Because the universe has expanded by a factor of about 1,100 since then, and expansion stretches the wavelength of every photon by that same factor. A blackbody whose every wavelength is stretched uniformly stays a blackbody, just a cooler one. The temperature scales inversely with the stretch.

$$T_\text{now} = \frac{T_\text{emit}}{1 + z} \approx \frac{3000\ \text{K}}{1090} \approx 2.725\ \text{K}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

This is the single most common thing to get wrong about the CMB. It is not radiation from cold matter. It is radiation from hot matter, emitted at 3,000 K, and then redshifted by a thousandfold journey through expanding space until it arrives as microwaves. The light didn't cool by losing heat to its surroundings. Space itself stretched underneath it.

CMB temperature

2.72548 K

±0.00057 K, the most perfect blackbody known

Age at emission

~380,000 yr

at redshift z ≈ 1090, when the fog cleared

Ripple amplitude

ΔT/T ≈ 10⁻⁵

about 30 microkelvin about the mean

The ripples are sound

Now to the patches. For a long time everyone expected the CMB to carry tiny temperature variations, because the universe today is lumpy: galaxies, clusters, vast empty voids. That structure had to grow from somewhere, from small seeds of overdensity in the early universe. Gravity would amplify the seeds over billions of years. So the seeds should be visible as faint hot and cold mottling on the last-scattering surface. COBE found them in 1992, at the level of about 30 microkelvin, confirming $\Delta T / T \approx 10^{-5}$ and earning George Smoot and John Mather a second Nobel for the mission. But the mottling is not random static. It has a very specific structure, and the structure is the interesting part.

Go back to before recombination. The plasma was a tight coupling of two ingredients: baryons (the protons and electrons, which have mass and gravity) and photons (which have enormous pressure). Picture an overdense patch. Gravity pulls baryons inward, compressing the patch. But compressing photons raises their pressure, and that pressure pushes back out. In, out, in, out. The patch oscillates. Across the whole universe, every overdense region is bouncing between gravitational infall and radiation pressure. The plasma is ringing like a struck bell.

These oscillations are genuine sound waves, pressure waves in a medium, and they travel at a specific speed set by how stiff the photon-baryon fluid is.

$$c_s = c\,\sqrt{\frac{1}{3(1+R)}} \approx \frac{c}{\sqrt{3}} \approx 1.7 \times 10^{5}\ \text{km/s}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

Here $R$ is the ratio of baryon to photon energy density; in the early plasma it is small, so $c_s$ sits close to $c/\sqrt{3}$, roughly 170,000 kilometres per second. That is the speed of sound in the early universe: over half the speed of light. These were not gentle ripples on a pond. They were relativistic shock-fast pressure waves screaming through a fluid hotter than the centre of a star.

The hot and cold spots are not warm and cool regions of gas. They are snapshots of standing sound waves, caught mid-vibration at the instant the universe stopped being able to carry sound at all.

Then recombination happened, and the music stopped. The moment the electrons bound up into neutral hydrogen, the photons decoupled from the baryons and streamed away. Without photon pressure, the restoring force vanished. The sound waves froze. Whatever phase each wave happened to be caught in (fully compressed, fully rarefied, somewhere between) is the phase it kept forever, printed onto the photons that left at that instant. The CMB temperature map is a photograph of standing waves, stilled at the click of the shutter.

Why the spots are a relief map, not a heat map

Here is the second thing almost everyone gets wrong, and it is subtle. The hot spots are not hotter and the cold spots are not colder, not really. Every patch of the last-scattering surface emits blackbody radiation at essentially the same 3,000 K. So why do they look different by the time they reach us?

Because the photons have to climb out of gravitational wells of different depth. A dense patch is a deeper well. A photon leaving a deep well loses energy fighting its way out, which redshifts it, which makes it look colder. So the densest patches, the ones that will eventually grow into galaxy clusters, appear as cold spots in the map. Underdense patches sit in shallow wells, lose less energy escaping, and look warmer. This is the Sachs-Wolfe effect: the temperature you measure is the local emission temperature shifted by the gravitational potential the photon had to escape.

$$\frac{\Delta T}{T} = \frac{1}{3}\,\frac{\Phi}{c^2}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

So the CMB map is a relief map of the early universe's density, drawn in the currency of gravitational potential, not a thermometer reading of hot and cold gas. A "cold spot" is a flag planted on an overdensity. Read the map that way and it inverts: the dark patches are where the matter was piling up, the bright patches are where it was thinning out. The cosmic web of galaxies and voids we live in today is the grown-up version of this exact pattern.

Reading the recipe off the power spectrum

The single most powerful thing you can do with the CMB is stop looking at individual spots and ask a statistical question instead: at what angular size are the spots most prominent? Take the whole map and decompose it into ripples of every scale, the way you'd break a chord into its constituent notes. Plot how much power sits at each angular scale, and you get the CMB power spectrum, the curve under the lab above. It is not a smooth slope. It has a series of clear peaks.

Those peaks are the frozen sound waves, sorted by wavelength. Here is the elegant part. All these oscillations started at the same time, at the Big Bang, and all stopped at the same time, at recombination. So a wave's phase at the freeze depends only on how fast it oscillated. The wave that had just completed exactly one half-cycle, compressing to maximum density right as the music stopped, produces the strongest signal. Its wavelength is the sound horizon: the distance a sound wave could travel from the Big Bang to recombination.

$$r_s = \int_0^{t_\text{rec}} c_s(t)\,\frac{dt}{a(t)} \approx 150\ \text{Mpc}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

That comes out to about 150 megaparsecs in comoving distance, and it is the loudest note the early universe could play, the fundamental tone of a struck bell whose size was the whole observable cosmos. The first peak in the power spectrum is this note. The second peak is the wave that completed a full cycle (compressed, then rarefied to maximum) by recombination, the first overtone. The third is the next harmonic, and so on. The CMB power spectrum is a harmonic series, the resonant modes of the entire universe.

And the heights of those peaks read out the recipe. The exact positions and relative heights depend on how much baryonic matter there was (baryons add inertia, deepening the compressions and boosting odd peaks over even ones), how much total matter, how fast the universe was expanding. By fitting the observed peaks, Planck pinned down the universe's composition: about 5% ordinary matter, 27% dark matter, 68% dark energy. That entire inventory falls out of measuring where the bumps sit in a curve. When you drag the baryon control in the lab and watch the peaks heave up and down, you are doing the thing cosmologists do for a living: turning the dials until the predicted music matches the recording.

What the baby photo actually tells us

Step back and look at what this featureless glow has given us. The temperature, 2.72548 K, is a thermometer reading on the universe at 380,000 years old, redshifted into our instruments. The blackbody perfection certifies that the early universe really was a hot dense equilibrium, the Big Bang picture rather than some steady alternative. The one-part-in-100,000 ripples are the seeds of every galaxy, caught the instant before gravity began amplifying them. And the harmonic series of acoustic peaks weighs out the contents of the cosmos to a few per cent, ordinary matter against dark matter against dark energy, from nothing more than the pattern of frozen sound.

There is something almost unreasonable about how much survives in this signal. A pressure wave bounced through a plasma for 400,000 years, got caught mid-oscillation, and then coasted untouched for 13.8 billion years across an expanding universe, and we can still read its phase off a satellite map. The early universe rang like a bell, the bell went silent at recombination, and the last reverberation is still arriving, stretched into microwaves, faint enough to mistake for pigeon droppings, carrying the recipe of everything. Some food for thought, the next time you tune a radio between stations and catch a hiss: a fraction of that static is the universe's baby photo, 13.8 billion years in the post.

Reading further

• Penzias & Wilson, A Measurement of Excess Antenna Temperature at 4080 Mc/s (ApJ 142, 1965). The accidental discovery paper, barely two pages; the understated tone is half the charm given what they'd found.
• Mather et al., Measurement of the CMB Spectrum by the COBE FIRAS Instrument (ApJ 420, 1994). The blackbody measurement that drew a standing ovation: deviations from Planck under 0.005%.
• Planck Collaboration, Planck 2018 results V: CMB power spectra and likelihoods (A&A 641, 2020). The definitive modern measurement of the acoustic peaks and the cosmic recipe they encode.
• Ryden, Introduction to Cosmology (2nd ed., Cambridge, 2016), ch. 9. The clearest textbook treatment of recombination, the sound horizon, and how the power spectrum turns into cosmological parameters.