Abstract. Accurate knowledge of the rate as well as the mechanism of excitation of the bending mode of CO₂ is necessary for reliable modeling of the mesosphere–lower thermosphere (MLT) region of the atmosphere. Assuming the excitation mechanism to be thermal collisions with atomic oxygen, the rate coefficient derived from the observed 15 μm emission by space-based experiments (k_ATM = 6.0 × 10⁻¹² cm³s⁻¹) differs from the laboratory measurements (k_LAB =(1.5-2.5) × 10⁻¹² cm³s⁻¹) by a factor of 2–4. The general circulation models (GCMs) of Earth, Venus, and Mars have chosen to use a median value of k_GCM = 3.0 × 10⁻¹² cm³s⁻¹ for this rate coefficient. As a first step to resolve the discrepancies between the three rate coefficients, we attempt to find the source of disagreement between the first two. It is pointed out that a large magnitude of the difference between these two rate coefficients (k_x ≡ k_ATM - k_LAB) requires that the unknown mechanism involve one or both major species: N₂, O. Because of the rapidly decreasing volume mixing ratio (VMR) of CO₂ with altitude, the exciting partner must be long lived and transfer energy efficiently. It is shown that thermal collisions with N₂, mediated by a near-resonant rotation-to-vibration (RV) energy transfer process, while giving a reasonable rate coefficient k_VR for de-excitation of the bending mode of CO₂, lead to vibration-to-translation k_VT rate coefficients in the terrestrial atmosphere that are 1–2 orders of magnitude larger than those observed in the laboratory. It is pointed out that the efficient near-resonant rotation-to-vibration (RV) energy transfer process has a chance of being the unknown mechanism if very high rotational levels of N₂, produced by the reaction of N and NO and other collisional processes, have a super-thermal population and are long lived. Since atomic oxygen plays a critical role in the mechanisms discussed here, it suggested that its density be determined experimentally by ground- and space-based Raman lidars proposed earlier.