The Cosmic Microwave Background
The oldest light in the universe arrived in New Jersey in 1964 disguised as a fault in the equipment.

In the spring of 1964 two Bell Labs radio astronomers, Arno Penzias and Robert Wilson, aimed a twenty-foot horn-reflector antenna at the sky above Holmdel, New Jersey, and picked up a hiss they could not account for. The instrument had not been built for astronomy. A. B. Crawford’s group put it up in 1960 to catch signals bounced off the Echo balloon satellite, and when the satellite work finished in 1963 the two of them inherited twenty-five square metres of collecting area with one unusual virtue. Its sensitivity could be calculated rather than estimated, and the horn kept ground radiation to less than a three-thousandth of what an open dish would admit. That made it a rare thing, an instrument for measuring absolute brightness, and what it measured, once everything known had been subtracted, was about three and a half degrees too warm.
They spent the better part of a year trying to get rid of the excess. They compared the antenna against a reference source cooled in liquid helium, ruled out New York City by pointing away from it, and ruled out the atmosphere by scanning from the zenith down toward the horizon and watching the signal fail to grow the way a thickening column of air would make it grow. A pair of pigeons had nested in the throat of the horn and coated the inside with what Wilson, in his Nobel lecture, called a white material familiar to all city dwellers; the birds were evicted, the mess cleaned out, and the temperature came down a little. A local source varies with something, the hour or the weather or the direction of the city. Theirs varied with nothing.

The explanation arrived sideways, in a conversation about something else entirely. Penzias mentioned the unexplained noise to Bernard Burke of MIT, who remembered hearing of theoretical work by Jim Peebles in Robert Dicke’s group at Princeton, thirty miles up the road. Penzias telephoned Dicke, who sent the preprint over. The Princeton group had argued that a universe which began hot would still be sitting in the cooled remains of that heat, Peebles had put the floor at ten degrees, and Dicke, Peter Roll and David Wilkinson were already building a small antenna of their own to hunt for it. Both papers ran in the Astrophysical Journal of 1 July 1965, the Princeton interpretation at page 414 and the Holmdel measurement at page 419. The measurement runs three pages, and its title promises nothing whatever.
Measurements of the effective of the 20-foot horn-reflector antenna at 4080 have yielded a value of about 3.5 K higher than expected. This excess temperature is, within the limits of our observations, , , and free from seasonal variations (July, 1964 – April, 1965).
A. A. Penzias and R. W. Wilson, “A Measurement of Excess Antenna Temperature at 4080 Mc/s,” The Astrophysical Journal 142 (1 July 1965), 419–421.
What they had walked into is the moment the universe stopped being opaque. For roughly its first 375,000 years, by NASA’s reckoning, the cosmos was hot enough to hold matter as a plasma of loose electrons and bare nuclei, and light could not cross it; photons scattered off the free electrons the way sunlight scatters inside fog. Expansion cooled the plasma until electrons could settle onto nuclei and make neutral atoms, the haze cleared, and the light already in the room flew free. It has been flying ever since. Expansion stretched its wavelengths by a factor of about a thousand along the way, from the glare of something white-hot down into microwaves, and the bath now sits at 2.72548 kelvin, plus or minus 0.00057.
The shape of the spectrum is what closes the argument. Anything in thermal equilibrium radiates along one curve, fixed entirely by its temperature. COBE, launched in November 1989, carried a spectrometer called FIRAS that set the background against that prediction. The points landed on the line so precisely that the error bars disappear inside its width. No assembly of stars, dust or radio galaxies emits that shape. Something has to have been in thermal contact with everything else, and the only epoch dense enough to allow it is the first one. John Mather and George Smoot took the 2006 Nobel for the blackbody form and for the structure found inside it.
That structure carries as much as the smoothness does. The sky is not perfectly even, and on NASA’s own description of the WMAP map one patch reads 2.7251 kelvin while another reads 2.7249. Those two ten-thousandths of a degree are the density ripples in the primordial plasma, and everything that has condensed since, every galaxy and star and planet and reader, grew out of the heavier ones. Measuring the ripple pattern precisely is how cosmology turned into an arithmetic subject. Nine years of WMAP data date the universe at 13.77 billion years, to within half a percent. The same ripples divide it into roughly five percent ordinary atoms, twenty-five percent dark matter and seventy percent dark energy.
Steady-state cosmology was the serious rival until 1965 and had no way to produce a thermal bath at three degrees. Its leading defender, Fred Hoyle, is the man who gave the other side its name, on the BBC Third Programme on 28 March 1949; the familiar story that he meant it as a sneer is contested, and Hoyle said later he had been reaching for an image his listeners could hold. The background remains the strongest single piece of evidence for a hot beginning, and the reason it holds up is unglamorous. It was found by two engineers who did not want it, with an instrument built for telephone calls, and it took them a year of careful subtraction to run out of ways to make it go away.