The cosmic microwave background has a temperature of 2.725 Kelvin today. The standard model predicts that this temperature scales linearly with redshift: at redshift z, the temperature should be 2.725(1+z) Kelvin. At redshift 0.68, it should be 4.58 Kelvin. This prediction follows directly from the adiabatic expansion of a thermal radiation field in an expanding universe — the photons redshift, the distribution remains Planckian, and the temperature drops inversely with the scale factor. The prediction has been tested at a few discrete redshifts using the Sunyaev-Zeldovich effect, which probes the CMB indirectly through its interaction with hot gas in galaxy clusters. But a direct measurement of the radiation temperature at a cosmological distance requires finding a physical system whose properties depend sensitively on the ambient radiation field.
Klimenko, Neeleman, and Balashev (arXiv 2602.22399, February 2026) make this measurement using molecular absorption lines in a quasar spectrum, determining the CMB temperature at z = 0.68 for the first time at intermediate redshift.
The method uses carbon monoxide molecules in an intervening gas cloud along the line of sight to a background quasar. CO molecules absorb quasar light at specific wavelengths corresponding to transitions between rotational energy levels. The populations of these rotational levels depend on the excitation conditions — collisions with hydrogen atoms, radiative pumping by local sources, and thermal equilibrium with the ambient radiation field. At the low densities of intergalactic absorbers, collisional excitation is minimal. If local radiation sources are also weak, the rotational level populations are set primarily by the CMB — the molecules act as thermometers measuring the radiation bath they sit in.
The rotational temperature extracted from the CO absorption pattern gives the CMB temperature at the cloud's redshift. The measurement at z = 0.68 yields T = 4.58 (+0.13/-0.17) Kelvin, consistent with the predicted value of 4.58 Kelvin from the standard T(z) = T0(1+z) scaling. The error bars are modest — roughly 3% — limited by the signal-to-noise of the quasar spectrum and systematic uncertainties in the excitation analysis.
The result is a confirmation, not a surprise. The standard prediction holds. But the measurement technique itself is remarkable — using individual molecules at cosmological distances as thermometers for the radiation that fills the universe. The molecules were not placed there for this purpose. They happen to sit in a gas cloud that happens to lie along the line of sight to a bright quasar, at a redshift where CO transitions are observable from the ground. The thermometer is an accident of geometry.
The physics is clean because the measurement is local to the absorber. The CO molecules respond to the radiation field at their location — not to the CMB as processed by cluster gas or other intermediate physics. The temperature they report is the temperature of the radiation bath at z = 0.68, directly. The adiabatic cooling law of the universe, checked molecule by molecule, sightline by sightline, at whatever redshifts the accidental thermometers happen to live.