Big-Bang cosmology predicts that the early Universe, once hot and dense, cools through expansion into the present-day 2.7 K cosmic microwave background (CMB). Whether the CMB spectrum is an ideal black body and whether its temperature varies slightly with direction are critical tests. Mather led the FIRAS instrument on COBE and confirmed the black-body spectrum with unprecedented accuracy. Smoot headed the DMR instrument and, for the first time, detected µK-level temperature fluctuations on 7-degree angular scales. These results underpin theories of galaxy formation and cosmic inflation and transformed observational cosmology into a precision science.

COBE carried three instruments. FIRAS, a 4 K-cooled Michelson interferometer, measured the CMB spectrum with 0.5 cm⁻¹ resolution and demonstrated agreement with a 2.725 ± 0.002 K black-body curve to better than 10⁻⁵. DMR employed differential receivers at 31, 53 and 90 GHz; after Galactic foreground subtraction it extracted ΔT/T ≈ 10⁻⁵ anisotropies beyond the quadrupole. The resulting angular correlation supported a globally flat Universe and a cold dark matter-dominated model, constraining the amplitude of primordial scalar perturbations. The detected “hot” and “cold” spots were interpreted as direct evidence of gravitational instability seeding matter clumping. These achievements guided the designs of WMAP and Planck, paving the way to the ΛCDM concordance model.

FIRAS alternated observations of the sky and a 4 K internal calibrator through a single-mode waveguide; differential spectra were analyzed in lag space, suppressing systematics below 10⁻⁵. The residual set an upper limit |µ| < 9 × 10⁻⁵ on µ-type and y-type spectral distortions, tightly constraining quasi-thermalization scenarios. The 7.2° DMR anisotropy maps were resampled to HEALPix Nside = 16; after applying a |b| > 20° Galactic cut, the spherical harmonics coefficients aℓm were maximum-likelihood estimated. The ℓ = 2–20 power spectrum Cℓ matched an adiabatic CDM model with n_s ≈ 1, Ω_0 ≈ 1, Ω_b ≈ 0.05 and h ≈ 0.5. Statistical isotropy tests (χ², Kolmogorov) ruled out systematic origins of the signal. COBE validated the BBKS transfer function and provided the σ₈ ≈ 1 normalization benchmark for later large-scale structure studies.