Resolving the mesospheric nighttime 4 . 3 μ m emission puzzle : comparison of the CO 2 ( ν 3 ) and OH ( ν ) emission models

In the 1970s, the mechanism of vibrational energy transfer from chemically produced OH(ν) in the nighttime mesosphere to the CO2(ν3) vibration, OH(ν)⇒ N2(ν)⇒ CO2(ν3), was proposed. In later studies it was shown that this “direct” mechanism for simulated nighttime 4.3 μm emissions of the mesosphere is not sufficient to explain space observations. In order to better simulate these observations, an additional enhancement is needed that would be equivalent to the production of 2.8–3 N2(1) molecules instead of one N2(1) molecule in each quenching reaction of OH(ν)+N2(0). Recently a new “indirect” channel of the OH(ν) energy transfer to N2(ν) vibrations, OH(ν)⇒ O(1D)⇒ N2(ν), was suggested and then confirmed in a laboratory experiment, where its rate for OH(ν = 9)+O(3P) was measured. We studied in detail the impact of the “direct” and “indirect” mechanisms on CO2(ν3) and OH(ν) vibrational level populations and emissions. We also compared our calculations with (a) the SABER/TIMED nighttime 4.3 μm CO2 and OH 1.6 and 2.0 μm limb radiances of the mesosphere–lower thermosphere (MLT) and (b) with groundand space-based observations of OH(ν) densities in the nighttime mesosphere. We found that the new “indirect” channel provides a strong enhancement of the 4.3 μm CO2 emission, which is comparable to that obtained with the “direct” mechanism alone but assuming an efficiency that is 3 times higher. The model based on the “indirect” channel also produces OH(ν) density distributions which are in good agreement with both SABER limb OH emission observations and ground and space measurements. This is, however, not true for the model which relies on the “direct” mechanism alone. This discrepancy is caused by the lack of an efficient redistribution of the OH(ν) energy from higher vibrational levels emitting at 2.0 μm to lower levels emitting at 1.6 μm. In contrast, the new “indirect” mechanism efficiently removes at least five quanta in each OH(ν ≥ 5)+O(3P) collision and provides the OH(ν) distributions which agree with both SABER limb OH emission observations and groundand space-based OH(ν) density measurements. This analysis suggests that the important mechanism of the OH(ν) vibrational energy relaxation in the nighttime MLT, which was missing in the emission models of this atmospheric layer, has been finally identified. Published by Copernicus Publications on behalf of the European Geosciences Union. 9752 P. A. Panka et al.: The 4.3 μm nighttime emission puzzle


Introduction
A detailed study of nighttime 4.3 µm emissions was conducted by López-Puertas et al. (2004) aimed at determining the dominant mechanisms of exciting CO 2 (ν 3 ), where ν 3 is the asymmetric stretch mode that emits 4.3 µm radiation.The nighttime measurements of SABER channels 7 (4.3 µm), 8 (2.0 µm), and 9 (1.6 µm) for geomagnetically quiet conditions were analyzed, where channels 8 and 9 are sensitive to the OH (ν ≤ 9) overtone radiation from levels ν = 8-9 and ν = 3-5, respectively.López-Puertas et al. (2004) showed a positive correlation between 4.3 µm and both OH channel radiances at a tangent height of 85 km.This correlation was associated with the transfer (Kumer et al., 1978) of energy of the vibrationally excited OH(ν) produced in the following chemical reaction (hereafter "direct" mechanism): and then further to CO 2 (ν 3 ) vibrations N 2 (1) + CO 2 (0) ↔ N 2 (0) + CO 2 (ν 3 ).(R3) However, López-Puertas et al. (2004) showed that calculations based on the Kumer et al. (1978) model do not reproduce the 4.3 µm radiances observed by SABER.Although accounting for energy transfer from OH(ν) did provide a substantial enhancement to 4.3 µm emission, a 40 % difference between simulated and observed radiance remained (for the SABER scan 22, orbit 01264, 77 • N, 3 March 2002, which was studied in detail) for altitudes above 70 km.In order to reproduce measurements these authors found that, on average, 2.8-3 N 2 (1) molecules (instead of Kumer's suggested value of 1) are needed to be produced after each quenching of OH(ν) molecule in Reaction (R2).Alternative excitation mechanisms that were theorized to enhance the 4.3 µm radiance (i.e., via O 2 and direct energy transfer from OH to CO 2 ) were tested but found to be insignificant.Recently, Sharma et al. (2015) suggested a new "indirect" mechanism of the OH vibrational energy transfer to N 2 , i.e., OH(ν) ⇒ O( 1 D) ⇒ N 2 (ν).Accounting for this mechanism, but only considering OH(ν = 9), these authors performed simple model calculations to validate its potential for enhancing mesospheric nighttime 4.3 µm emission from CO 2 .They reported a simulated radiance enhancement between 18 and 55 % throughout the MLT.In a latest study, Kalogerakis et al. (2016) provided a definitive laboratory confirmation for the validity of this new mechanism and measured its rates for OH(ν = 9) + O.
We studied in detail the impact of "direct" and "indirect" mechanisms on the CO 2 (ν 3 ) and OH(ν) vibrational level populations and emissions and compared our calculations with (a) the SABER/TIMED nighttime 4.3 µm CO 2 and OH 1.6 and 2.0 µm limb radiances of MLT and (b) with the ground and space observations of the OH(ν) densities in nighttime mesosphere.
The study was performed for quiet (non-auroral) nighttime conditions to avoid accounting for interactions between charged particles and molecules, whose mechanisms still remain poorly understood.

Non-LTE model
A non-LTE (non-local thermodynamical equilibrium) analysis was applied to CO 2 and OH using the non-LTE ALI-ARMS (Accelerated Lambda Iterations for Atmospheric Radiation and Molecular Spectra) code package (Kutepov et al., 1998;Gusev and Kutepov, 2003;Feofilov and Kutepov, 2012), which is based on the accelerated lambda iteration approach (Rybicki and Hummer, 1991).
Our CO 2 non-LTE model is described in detail by Feofilov and Kutepov (2012).We modified its nighttime version to account for the "direct" mechanism, Reactions (R1)-(R3), in a way consistent with that of López-Puertas et al. ( 2004) and added the "indirect" mechanism of Sharma et al. (2015) and Kalogerakis et al. (2016) as described in detail below.Our OH non-LTE model resembles that of Xu et al. (2012).et al. (2015) suggested an additional mechanism that contributes to the CO 2 (ν 3 ) excitation at nighttime, and discussed in detail its available experimental and theoretical evidence.According to this mechanism, highly vibrationally excited OH(ν), produced by Reaction (R1), rapidly loses several vibrational quanta in collisions with O( 3 P) through a fast, spin-allowed, vibration-to-electronic energy transfer process that produces O( 1 D):

Sharma
Recently, Kalogerakis et al. (2016) presented the first laboratory demonstration of this new OH(ν) + O( 3 P) relaxation pathway and measured its rate coefficient for ν = 9.
The production at nighttime of electronically excited O( 1 D) atoms in Reaction (R4) triggers well-known pumping mechanism of the 4.3 µm emission, which was studied in detail for daytime (Nebel et al., 1994;Edwards et al., 1996).
Here O( 1 D) atoms are first quenched by collisions with N 2 in a fast spin-forbidden energy transfer process: then N 2 (ν) transfers its energy to ground state N 2 via a very fast single-quantum VV (vibrational-vibrational) process: leaving N 2 molecules with an average of 2.2 vibrational quanta, which is then followed by Reaction (R3).

Collisional rate coefficients
We use, in our CO 2 non-LTE model, the same VT (vibrational-translational) and VV collisional rate coefficients for the CO 2 lower vibrational levels as those of López-Puertas et al. (2004).However, a different scaling of these basic rates is applied for higher vibrational levels using the first-order perturbation theory as suggested by Shved et al. (1998).
The reaction rate coefficients applied in this study for modeling OH(ν) relaxation transfer of OH(ν) vibrational energy to the CO 2 (ν 3 ) mode are displayed in Table 1.The total chemical production rate of OH(ν) in Reaction (R1) was taken from Sander et al. (2011) and the associated branching ratios for ν were taken from Adler-Golden (1997).We treat Reaction (R2) both as a single (1Q, ν = 1) and multiquantum (MQ, ν = 2 or 3) quenching process.We use the rate coefficient of this reaction (with associated branching ratios) taken from Table 1 of Adler-Golden (1997) and multiplied it by a low temperature factor of 1.4 (Lacoursière et al., 2003) for MLT regions.The rate coefficient for Reaction (R3) was taken from Shved et al. (1998).

Model inputs and calculation scenarios
The nighttime atmospheric pressure, temperature, and densities of trace gases and main atmospheric constituents for calculations presented below were taken from the WACCM (Whole Atmosphere Community Climate Model) model (Marsh et al., 2013).
The following sets of processes and rate coefficients were used in our calculations: This is our basic model version with both "direct", Reaction (R2), and "indirect", Reaction (R4) + (R5), mechanisms working together when Reaction (R2) is treated as the single-quantum process (ν = 1) as was suggested by Kumer et al. (1978), though Reaction (R6) is treated as the multi-quantum process (any ν ≤ ν −1).
A new mechanism, Reactions (R4) and (R5), replaces here Reaction (R7), which is used in other models described above.

Vibrational temperatures of the CO 2 (ν 3 ) levels
The vibrational temperature T ν is defined from the Boltzmann formula which describes the excitation degree of level ν against the ground level 0.Here g ν and E ν are the statistical weight and the energy of level ν, respectively.If T ν = T kin then level ν is in LTE.
Figure 1 shows the vibrational temperatures of the CO 2 levels of four isotopes, giving origin to 4.3 µm bands, which dominate the SABER nighttime signal (López-Puertas et al., 2004).These results were obtained for SABER scan 22, orbit 01264, 77 • N, 3 March 2002.The same scan was used for the detailed analysis presented in the work by López-Puertas et al. (2004).The kinetic temperature retrieved for this scan from the SABER 15 µm radiances (SABER data version 2.0) and vibrational temperature of N 2 (1) are also shown.Solid lines in Fig. 1 represent simulations with our basic model [(OH-N2 1Q) & (OH-O2 MQ) & Reactions (R4), (R5)], when both the "direct" process Reaction (R2) (in its singlequantum version, as was suggested by Kumer et al., 1978), and the new "indirect" process, Reactions (R4) + (R5) are included.For comparison we also show vibrational temperatures (dashed lines) for the model [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)], where the "indirect" mechanism is off and the "direct" process is treated as a three-quantum one, which is equivalent to the 3 times higher efficiency suggested by López-Puertas et al. (2004).
In both simulations, CO 2 (00011) of main isotope 626 and N 2 (1) have almost identical vibrational temperatures up to ∼ 87 km which is caused by an efficient VV exchange (Reaction R3).(1997), as a multi-quantum VV process.Compared to previous model this run shows a significantly lower channel 7 signal.This is obviously caused by a much more efficient removal of the OH(ν) vibrational energy in the multi-quantum quenching by collisions with O 2 .As a result, a significantly smaller part of this energy is collected by N 2 (1) and delivered to the CO 2 (ν 3 ) vibrations with the "direct" mechanism Reaction (R2).To compensate this OH(ν) decay and keep the transfer of energy to CO 2 (ν 3 ) unchanged, López-Puertas et al. ( 2004) adjusted new, presumably higher OH(ν) to the SABER 1.6 and 2.0 µm radiances.In our study the higher channel 7 emission is, however, restored when we include the Reactions (R4) and (R5) ("indirect" mechanism of energy transfer from OH(ν) to CO 2 (ν 3 )) in the model, but return Reaction (R2) to its single-quantum mode.
In Fig. 2, we show (black curve with diamonds) the channel 7 radiance profile for the SABER scan specified in Sect.3.1.Comparing turquoise and red curves with this measurement, one may see that both the "direct" mechanism alone in its three-quantum version and the combination of "indirect" and single-quantum "direct" mechanisms are close to the SABER radiance for this scan.However, to provide this pumping level, the multi-quantum "direct" mechanism needs to be supported by the inefficient single-quantum OH(ν) quenching in Reaction (R6) by collisions with O 2 , which helps maintain a higher population of OH(ν).We also note here that both our violet  2004), short-dash and solid lines, respectively, very well.
We also show in Fig. 2 our study of how both the "direct" and "indirect" mechanisms work together when the "direct" process Reaction (R2) is treated as a multi-quantum.The magenta curve in this plot is the result obtained with the model [(OH-N2 3Q) & (OH-O2 MQ) & Reactions (R4), (R5)] when Reaction (R2) is treated as a three-quantum process.This combination of both mechanisms provides high CO 2 (ν 3 ) pumping and subsequently strong channel 7 emission.The latter exceeds the turquoise and red curves by 20-45 % in the altitude range considered and strongly deviates from the measured radiance profile.
Two other results of this study are shown only on the right panel of this plot for the signal differences.The light blue curve corresponds to simulations with the [(OH-N2 2Q) & (OH-O2 MQ) & Reactions (R4), (R5)] model when the quantum transfer in Reaction (R2) is reduced from 3 to 2. The dark green curve is obtained for the case when Reaction (R2) is treated as two-quantum process for highly resonant transitions OH(9) + N 2 (0) → OH(7) + N 2 (2) and OH(2) + N 2 (0) → OH(0) + N 2 (2), and as single-quantum for all other vibrational levels.It is seen that both of these input versions bring the calculated radiance closer to our result for a singlequantum "direct" process Reaction (R2) (red curve  One may see that in Fig. 3 the "direct" mechanism alone with three-quantum efficiency for Reaction (R2), as well as both the "direct" (as single-quantum) and "indirect" mechanisms together provide similar results for all four atmospheric models, within a 10 to 30 % difference range.By comparing these models to measured radiances, both calculations are close to the observed signal down to 68 km for MLW and down to 75 km for SAW.For MLS and TROP, the two-mechanism calculations are somewhat closer to measurements than those for "direct" mechanism alone in altitude interval 75-90 km.

Comparison of OH vibrational populations with ground-and space-based observations
In Fig. 4 we present relative OH(ν) populations calculated using three different sets of rate coefficients discussed in the previous section, which provided comparably high enhancement of the CO 2 (ν 3 ) emission.These calculations are compared with the vibrational populations derived from ground (panel a) and space-based (panel b) observations of OH emissions.
Measured populations (black) displayed in panel (a) were recorded by Cosby and Slanger (2007)  To achieve a high signal-to-noise ratio, 300 radiance spectra (OH ν = 1 and 2) were collected and averaged.We show in panel (b) of Fig. 4 the OH(ν) population distribution normalized to OH(ν = 9) derived in this study as well as corresponding uncertainties.The three simulated distributions (red, turquoise, and magenta) in this panel are modeled using WACCM inputs taken on 22 November 2000 at latitude 45 • N at local midnight.
To simulate the ground-based observations of Cosby and Slanger (2007), panel (a), the calculated relative populations were integrated over the entire altitude range of our model (30-135 km).For panel (b), we have integrated calculated OH(ν) densities of 87 to 105 km as observed by VIRTIS from Migliorini et al. (2015) to simulate mean population distribution obtained in this study.
The turquoise profiles in both panels of Fig. 4 represent results obtained with our set of rate coefficients [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)] similar to the one used in López-Puertas et al. (2004), where the authors treated the OH-N 2 reaction with an efficiency increased by a factor of 3, the OH-O 2 reaction as single-quantum, and the OH-O reaction as a "sudden death" quenching or chemical OH removal process (Reaction R7), with ν independent rate coefficient of 2.0 × 10 −10 cm 3 s −1 .In panel (a), the turquoise profile shows higher relative populations compared to measurements for upper vibrational levels ν > 4, whereas in panel (b) this model shows populations within the uncertainty range of measurements for ν > 4. For lower vibrational levels ν ≤ 4, the populations calculated with this model are, however, significantly lower than measured ones: by 30 % for ν = 3 in the panel (a) and by up to 85 % for ν = 1 in panel (b).A significantly slower increase in populations calculated with the [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)] model compared to measurements can be explained by the lack of efficient mechanisms redistributing the OH(ν) energy from higher vibrations levels to lower ones.The single-quantum OH-O 2 reaction also allows for more excited OH molecules in the upper vibrational levels relative to a multi-quantum process.Additionally, a slower increase in calculated populations with the ν decreasing compared to measured ones which is seen in both panels, is the effect of the high quenching rate coefficient of the Reaction (R7) for lower vibrational levels for which this reaction dominates over the singlequantum O 2 quenching.
The situation is different when our basic model [(OH-N 2 1Q) & (OH-O 2 MQ) & Reactions (R4), (R5)] is applied (red curves).As discussed above in the previous section, this model provides the same level of the CO 2 (ν 3 ) emission pumping as the extreme model of López-Puertas et al. (2004); compare red and turquoise curves in Fig. 2.However, they demonstrate significantly different population distributions.Relative OH population distribution in panel (a) shows our standard model in very good agreement with the results from Cosby and Slanger (2007), falling completely within the variation range of these measurements.Panel (b) also shows excellent agreement between calculations and measurements, where the former lie nearly completely within the measurement error bars for the majority of vibrational levels.In both panels our results reproduce well the steady upward trend in populations from upper to lower vibrational levels.Significantly higher populations of lower OH levels in this model are the result of redistribution of highervibrational-level energy to lower levels due to two dominant multi-quantum quenching mechanisms, namely the new Reaction (R4) and the multi-quantum version of Reaction (R6).We also note that Reaction (R4) uses a lower rate coefficient than Reaction (R7) for quenching the lower vibrational levels ν < 5, which results in maintaining their higher populations.
Measured OH(ν = 3) (panel b) was the only population which showed disagreement with our model.Various reasons of increased measured population at ν = 3 are discussed by Migliorini et al. (2015); however, no definitive conclusions were given.
Above 90 km atomic oxygen density increases rapidly with the altitude.As a result the role of Reaction (R4) in quenching higher OH vibrational and pumping lower levels increases.This effect is easily seen in panel (b) of Fig. 4, where mean OH(ν) densities for higher altitude region 87-105 km are compared.The turquoise curve (no Reaction R4) in this panel shows lower populations compared to those calculated with Reaction (R4) included.
The magenta profiles in both panels represent our calculations with the model [(OH-N 2 3Q) & (OH-O 2 MQ) & Reactions (R4), (R5)], which is identical to our standard model except for the Reaction (R2) treated as the three-quantum one.The multi-quantum OH-N 2 VV transfer provides faster quenching of excited OH, hence, a lower overall population of the magenta profiles compared to red profiles.Despite showing reasonable agreement with measurements in both panels, this model caused, however, an excessive increase for the 4.3 µm emissions, as seen in Fig. 2.
3.4 OH 1.6 and 2.0 µm emissions SABER channels 8 (2.0 µm), and 9 (1.6 µm) are dominated by the OH(ν) emission from levels ν = 8-9 and ν = 3-5, respectively.We simulated channel 8 and 9 radiances for four atmospheric models from Table 2. Results are shown in Fig. 3, bottom row, as ratios of volume emission rates for R2), in turquoise.Black curves in this plot display SABER-measured VER ratios, for which VERs were obtained with the Abel inversion procedure, similar to that described by López-Puertas et al. (2004), from the SABER channel 8 and 9 limb radiances for scans listed in Table 2.
Comparing red and turquoise profiles in Fig. 3 (bottom row), one may see that our standard model (red) shows significantly lower VER ratios for altitudes 85-100 km than the model of López-Puertas et al. (2004), turquoise.These differences between ratios are a result of very different OH(ν) population distributions (Fig. 4) for each model, which were discussed in the previous section.The channel 8/channel 9 VER ratios reflect these distributions very well since channel 8 is sensitive to the OH(ν) emissions from higher levels 8 and 9, whereas channel 9 records emissions of lower levels 3-5.A significantly higher population of lower vibrational levels in our model (red curves in Fig. 4) explain low VER ratios.In contrast, the model [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)], which underpredicts lower level populations, provides VER ratios which significantly exceed both our model results and measurements for altitudes above 90 km, where [O] density rapidly increases with altitude.This comparison demonstrates the strong impact of Reaction (R2), which provides efficient quenching of higher OH vibrational levels in collisions with O( 3 P) atoms in this altitude region.
4 Conclusions Kumer et al. (1978) first proposed the transfer of vibrational energy from chemically produced OH(ν) in the nighttime mesosphere to the CO 2 (ν 3 ) vibration, OH(ν) ⇒ N 2 (ν) ⇒ CO 2 (ν 3 ).The effect of this "direct" mechanism on the SABER nighttime 4.3 µm emission was studied in detail by López-Puertas et al. (2004), who showed that in order to match observations, an additional enhancement is needed that would be equivalent to the production of 2.8-3 N 2 (1) molecules instead of one molecule for each quenching reaction OH(ν) + N 2 (0).López-Puertas et al. (2004) concluded that the required 30 % efficiency in the OH(ν) + N 2 (0) energy transfer is, in principle, possible, although the mecha-nism(s) whereby the energy is transferred is (are) not currently known.
Recently, Sharma et al. (2015) suggested a new efficient "indirect" channel of the OH(ν) energy transfer to the N 2 (ν) vibrations, OH(ν) ⇒ O( 1 D) ⇒ N 2 (ν) and showed that it may provide an additional enhancement of the MLT nighttime 4.3 µm emission.Kalogerakis et al. (2016) provided a definitive laboratory confirmation of new OH(ν) + O vibrational relaxation pathway and measured its rate for OH(ν = 9) + O.
We included the new "indirect" energy transfer channel in our non-LTE model of the nighttime MLT emissions of CO 2 and OH molecules and studied in detail the impact of "direct" and "indirect" mechanisms on simulated vibrational level populations and radiances.The calculations were compared with (a) the SABER/TIMED nighttime 4.3 µm CO 2 and OH 1.6 and 2.0 µm limb radiances of MLT and (b) with the ground and space observations of the OH(ν) densities in the nighttime mesosphere.We found that new "indirect" channel provides significant enhancement of the 4.3 µm CO 2 emission.This model also produces OH(ν) density distributions which are in good agreement with both SABER limb OH emission measurements and the ground and space observations in the mesosphere.Similarly strong enhancement of 4.3 µm emission can also be achieved with the "direct" mechanism alone assuming a factor of 3 increase in efficiency, as was suggested by López-Puertas et al. (2004).This model does not, however, reproduce either the SABER-measured VER ratios of the OH 1.6 and 2.0 µm channels or the ground and space measurements of the OH(v) densities.This discrepancy is caused by the lack of efficient redistribution of the OH(ν) energy from the higher vibrational levels emitting at 2.0 µm to lower levels emitting at 1.6 µm in the models based on the "direct" mechanism alone.In contrast, this new "indirect" mechanism (Reactions R4 and R5 of Table 1), efficiently removes at least five quanta in each OH(ν) + O( 3 P) collision from high OH vibrational levels.Supported also by the multi-quantum OH(ν) + O 2 quenching (Reaction R6 of Table 1), the new mechanism provides OH(ν) distributions which are in agreement with both measured VER ratios and observed OH(ν) populations.
The results of our study suggest that the missing nighttime mechanism of CO 2 (ν 3 ) pumping has finally been identified.This confidence is based on the fact that the new mechanism accounts for most of the discrepancies between measured and calculated 4.3 µm emission for various atmospheric situations, leaving relatively little room for other processes, among them the multi-quantum "direct" mechanism.The accounting for the multi-quantum transfer in reaction OH(v) + N 2 together with the "indirect" mechanism has little influence on the OH(ν) population distributions; however, it can enhance the 4.3 µm emission.Therefore, further laboratory and/or theoretical investigation of this reaction is needed to define its role.Further improvements for the new "indirect" mechanism will require optimizing the set of rate coef-ficients used for OH(ν) relaxation by O( 3 P) and O 2 at mesospheric temperatures and, in particular, understanding the dependence of the "indirect" mechanism on the OH vibrational level.Relevant laboratory measurements and theoretical calculations are sorely needed to understand these relaxation rates and the quantitative details of the applicable mechanistic pathways.Nevertheless, the results presented here clearly demonstrate significant progress in understanding the mechanisms of the nighttime OH and CO 2 emission generation in MLT.

Figure 2
Figure 2 displays our simulations of SABER channel 7 (4.3 µm) radiances for inputs which correspond to the measurement conditions of the SABER scan described in Sect.3.1.The calculations also account for the minor contribution in channel 7 radiation emitted by the OH(ν ≤ 9) vibrational levels.Our simulation for this scan with the [(OH-N2 1Q) & (OH-O2 1Q) & Reaction (R7)] set of rate coefficients is shown by the violet curve.The turquoise curve displays our results for the rate coefficient set [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)], which simulates the model suggested by López-Puertas et al. (2004) with the factor of 3 increased efficiency of Reaction (R2).One may see that treating Reaction (R3) as a three-quantum VV process strongly enhances the pumping of the CO 2 (ν 3 ) vibrations and that the 4.3 µm radiance is in agreement with López-Puertas et al. (2004) results.The blue curve in Fig. 2 displays our run with the model [(OH-N2 3Q) & (OH-O2 MQ) & Reaction (R7)].In this model Reaction (R6) is treated, following Adler-Golden(1997), as a multi-quantum VV process.Compared to previous model this run shows a significantly lower channel 7 signal.This is obviously caused by a much more efficient removal of the OH(ν) vibrational energy in the multi-quantum quenching by collisions with O 2 .As a result, a significantly smaller part of this energy is collected by N 2 (1) and delivered to the CO 2 (ν 3 ) vibrations with the "direct" mechanism Reaction (R2).To compensate this OH(ν) decay and keep the transfer of energy to CO 2 (ν 3 ) unchanged, López-Puertas et al. (2004) adjusted new, presumably higher OH(ν) to the SABER 1.6 and 2.0 µm radiances.In our study the higher channel 7 emission is, however, restored when we include [(OH-N2 1Q) & (OH-O2 1Q) & Reaction (R7)] and turquoise [(OH-N2 3Q) & (OH-O2 1Q) & Reaction (R7)] curves reproduce the corresponding results in Fig. 10 of López-Puertas et al. (
on 3 March 2000 using the echelle spectrograph and imager (ESI) on the Keck II telescope at Mauna Kea (19.8206 • N, 155.4681 • W).The authors measured emission intensities of the 16 OH Meinel bands which were converted into the OH(ν) column densities and normalized to column density of OH(ν = 9).Several observations of OH emissions were recorded throughout the night.We display the average column densities as well as their variation ranges for each vibrational level.The three simulated distributions (red, turquoise, and magenta) in this panel are modeled using WACCM inputs taken on 3 March 2000 at latitude 20 • N at local midnight.Measured densities displayed in panel (b) of Fig. 4 were taken from Migliorini et al. (2015), who analyzed VIRTIS (for Visible and Infrared Thermal Imaging Spectrometer) measurements on board the Rosetta mission.VIRTIS performed two limb scans of the OH Meinel bands from 87 to 105 km covering the latitude range from 38 to 47 • N between 01:30 and 02:00 solar local time, in November 2009.

Table 1 .
Significant collisional processes used in model.