Transition 2₂₀ ← 2₂₁ at 23 GHz measured at a nominal trap temperature of 15 K. The broadness of this feature can be recognized by comparison with the narrow Gaussian curve, which is the shape expected for a single line at 40 K. Since pressure or Doppler broadening to this extent can be excluded, unresolved hyperfine structure is the most reasonable explanation for this broadening, whereas some residual power broadening effect cannot be excluded. These hyperfine splittings are dominated by the small electrical quadrupole moment of the deuterium nucleus. To better identify the structure in this feature, the data of 13 single measurements (shown as pale grey dots) have been normalized (to the off-resonant CH counts), averaged, and binned into 6 kHz steps. The result is shown as a red curve, where two smaller peaks become discernible on both sides of the feature.