The longitudinal variation of thermospheric O/N2 column density ratio is studied, based on the TIMED/GUVI measurements collected during 2002-2007. In October, a four-peaked longitudinal fluctuation in O/N2 is found. The wavenumber-4 fluctuation in equatorial O/N2 is generally stronger at lower solar activity, and maximizes in the morning of boreal summer and autumn. In boreal winter, the wavenumber-4 component gives way to the wavenumber-3. Such seasonal variation and solar activity dependence are consistent with those of the diurnal eastward-propagating mode with wavenumber-3, but the local time dependence is not. The daytime wavenumber-4 pattern in O/N2 does not shift eastward in local time as expected if it is directly connected to the DE3. No definite explanation is found for this structure, but a consideration of various mechanisms suggests that the vertical wind effect may be the most potential candidate.
 The diurnal cycle of solar insolation excites atmospheric tides characterized as large-scale waves in temperature, density and winds with periods that are harmonics of a solar day [e.g., Hagan and Forbes, 2002; Forbes et al., 2008, and references therein]. The non-sunsynchronous tidal components are termed as nonmigrating tides. Among the nonmigrating tides, the diurnal eastward-propagating mode with wavenumber-3 (DE3 for short) is the most dynamical component in the thermosphere during most of the year [e.g., Hagan and Forbes, 2002]. The DE3 mode is suggested to be induced by latent heat release in the troposphere [Hagan and Forbes, 2002]. When propagating upward, the DE3 mode grows exponentially. Finally, the tide dissipates and deposits momentum and energy directly into the mesosphere, lower thermosphere, and possibly the upper thermosphere [Hagan et al., 2009]. There is a general agreement that the DE3 mode imposes remarkable signatures in the ionosphere, thermosphere and mesosphere. When the DE3 tide is observed in a fixed local time frame such as from the slowly precessing satellite, four zonal wave patterns (i.e., wavenumber-4 structure, or simply WN4) would be recorded diurnally, due to the blueshift of Doppler Effect [also cf. Oberheide et al., 2003].
 In the ionospheric F-region, electron production is primarily due to the photoionization of atomic oxygen and the loss rate is proportional to the density of molecular compositions (primarily N2). As a result, the temporal and spatial variations of O/N2 play an important role in the changes of plasma density in the F-region. The primary purpose of the present work is to study the longitudinal O/N2 variation and search for tidal signatures in the O/N2 ratio. After a brief description about the dataset in Section 2, we present the longitudinal O/N2 structure and identify their possible origin in Section 3.
 The NASA's Thermosphere, Ionosphere, and Mesosphere, Energetics and Dynamics (TIMED) spacecraft orbits at 630 km altitude with an inclination of 74.1° [Christensen et al., 2003]. An instrument on board the TIMED is the Global Ultraviolet Imager (GUVI), which measures the brightness of the far ultraviolet (FUV) airglow in five bands: 121.6 nm, 130.4 nm, 135.6 nm, 141.0–152.8 nm, and 167.2–181.2 nm. It returns the images of the five “colors” with global coverage, which can be used to monitor the composition, temperature, large-scale wave structure, and auroral processes in the upper atmosphere. The column abundance of thermospheric atomic oxygen relative to molecular nitrogen is derived from the simultaneous measurements of disk-viewing dayglow of OI (135.6 nm) and N2 Lyman-Birge-Hopfield (LBH) [Strickland et al., 2004]. The derived O/N2 is the ratio of height-integrated O density to the height-integrated N2 density above a reference level near the bottom of the photoelectron-excited dayglow layer (a variable altitude of approximate 140 km, above which the height-integrated density of N2 is 1017 cm−2). Thermospheric O/N2 for this investigation is level 3 data obtained from the Johns Hopkins University Applied Physics Laboratory. We used the measurements within the period from day of year (DOY) 008, 2002 to DoY 307, 2007.
3. Results and Conclusions
Figures 1a–1d show the global maps of GUVI-measured O/N2 in four 10-day-wide windows around the equinoxes and solstices. In each date window, the 5 years of measurements collected during 07-09 LT under geomagnetically quiet condition are binned, and the median is computed for each grid with a resolution of 1.765° longitude by 1.765° latitude. The quiet condition is characterized by AE6 index < 300nT [Werner and Prölss, 1997] and Estimated Hemispheric Power < 30GW [e.g., Emery et al., 2008]. The most distinct feature in Figures 1a–1d is the asymmetry between the Northern and Southern Hemispheres, i.e., the O/N2 ratio is generally lower in the Hemisphere where the subsolar point locates at (the Southern Hemisphere in October and December and the Northern Hemisphere in April and June). This feature could be attributed to the transequatorial neutral wind driven by the latitudinal gradient of solar irradiance heating [Roble et al., 1977]. Meanwhile, the O/N2 ratio in Figures 1a–1d exhibit prominent longitudinal variations. In the northern and Southern Hemispheres, the O/N2 ratio is generally lower in the longitudinal sectors near the local geomagnetic pole (90°W in the Northern Hemisphere and 120°E in the Southern Hemisphere) than that in longitudinal sectors away from the pole. This longitudinal structure might have a contribution to the Weddell Sea Anomaly [He et al., 2009] and could arise from the fact that the north and south geomagnetic poles locate at different longitudes [e.g., Rishbeth and Müller-Wodarg, 1999; Lei et al., 2010]. Furthermore, a striking four-peaked longitudinal structure is clearly visible around September Equinox (Figure 1c). Between 20°S and 40°N geomagnetic latitudes, the O/N2 ratio peaks distinctly around 150°W, 50°W, 30°E and 120°E. In other seasons the zonal fluctuation can also be observed around the geomagnetic equator, although four-peaked structure is not clearly seen as that around September Equinox.
 To compare each component in the zonal O/N2 fluctuation, a Fourier analysis is performed for the Equatorial O/N2 Ratio (EON2R). The EON2R, the averaged O/N2 within ±20° geomagnetic latitudes, is calculated by the following equation with a longitudinal interval of 7.05°:
The Fourier amplitude spectra of the EON2R of Figures 1a–1d are exhibited in Figures 1e–1h, respectively. Generally, the significant components are wavenumber-1, 2, 3, and 4. The WN4 component presents a seasonal variation. It is strong around September Equinox, less strong around June Solstice and March Equinox, and weak around December Solstice, which is consistent with the seasonal variation of the WN4 signatures in total electron content (TEC), vertical plasma drifts, temperature, and zonal wind [Kil et al., 2008; Oberheide and Forbes, 2008b; Wan et al., 2008; Ren et al., 2009]. It is broadly believed that the DE3 tide excited in the tropical troposphere is the major contributor to the observed WN4 in various ionosphere-thermosphere properties [Hagan et al., 2007; Wan et al., 2008]. The DE3 mode is the most dominant component in all nonmigrating tidal modes for most of the year, but it experiences a considerable intraannual variation [e.g., Oberheide et al., 2006] and sometimes can be met or exceeded by the diurnal eastward-propagating mode with wavenumber-2 (DE2), [Häusler et al., 2007; Pedatella et al., 2008; Forbes et al., 2008, 2009; Häusler and Lühr, 2009]. Although the details of DE3 mode may vary with altitude, solar activity level or measuring method, there is a consensus that the DE3 mode maximizes during July–September, and it is dominant mainly during boreal summer and autumn but in an inferior position during boreal winter compared with the DE2 mode. When the DE2 dominates over the DE3, a zonal wavenumber-3 (WN3) structure could be observed in a fixed local time frame [Oberheide and Forbes, 2008a]. Such seasonal transition from WN4 to WN3 was observed in the zonal fluctuations of in-situ electron densities [Pedatella et al., 2008], nitric oxide densities [Oberheide and Forbes, 2008a], and other ionospheric and thermospheric properties. In Figure 1h, the amplitude of WN4 component declines around December Solstice while the WN3 component enhances, compared with that in the other seasons.
 GUVI's measurements in a 9-day-wide sliding window from DoY 008, 2002 to DoY 307, 2007 are binned to construct global O/N2 maps similar as those in Figures 1a–1d. In total, 1749 maps are constructed. The EON2R and its Fourier spectra for each map are calculated. The WN4 amplitudes of the series of EON2R spectrum are showed in Figure 2a as a function of LT and the 9 days mean of 10.7 cm solar radio flux (F10.7, in unit of sfu, 1sfu=10−22 W·m−2·Hz−1). Figure 2b shows the average of the WN4 amplitudes along F10.7. It is clear from Figures 2a and 2b that the WN4 amplitude is generally smaller at higher F10.7 level, which is consistent with the F10.7 dependence of DE3 mode [e.g., Oberheide et al., 2009], possibly resulting from the enhanced tidal dissipation at higher solar activity caused by the temperature dependence (proportional to T2/3) of molecular thermal conductivity [Oberheide et al., 2009]. Another notable feature in Figure 2a is the LT dependence. The LT dependence is studied by binning the WN4 amplitudes under F10.7<120 within six LT bands and four seasons (91 days around equinox or solstice). As shown in Figure 2c, the WN4 amplitude is most significant at 06–08 LT band during June Solstice and September Equinox; it declines during the following hours and turns to growth before 18 LT.
 If the DE3 tide dominates the WN4 pattern in O/N2, the pattern should be propagating eastward with a speed of about 90° per day observed in the LT-frame as the DE3-induced WN4 patterns in some other properties [e.g., Wan et al., 2008; Häusler and Lühr, 2009]. The WN4 phase distributions, for the six LT bands, of the series of EON2R spectrum are present in Figure 2d. The WN4 component peaks most probably between 21–33°E from 06–08 to 14–16 LT, and the longitude of maximum probability does not shift eastward with local time. Note that the lack of local time shift is not masked by the seasonal variation. When the data were binned for each season, the feature of local time shift is still not found (not shown). Thus, it suggests the WN4 pattern in O/N2 is not modulated directly by the DE3 tide in most of the daytime. Of all tide components contributing potentially to WN4 pattern, the stationary planetary with wavenumber-4 (sPW4) is the only stationary one. However, the sPW4 mode is significantly smaller than the DE3 mode [e.g., Häusler and Lühr, 2009; Hagan et al., 2009] and it has not been reported to be mainly responsible for any WN4 signatures. To satisfy the observational constraints, we speculate that the WN4 pattern in O/N2 might be excited by the DE3 tide at an earlier time, and then persists in local time.
 GUVI O/N2 ratio is derived theoretically from the brightness ratio of photoelectron-impact-excited OI 135.6nm emission to N2 LBH emission. In fact, the radiative recombination of oxygen ions and electrons will also produce 135.6nm emissions which are proportional to the product of the number densities of oxygen ion (NO+) and electron (Ne), approximately equal to Ne2 in the F-region. The remaining question is whether the radiative recombination could produce such a change that the WN4 fluctuation in the derived O/N2 is a signature of the fluctuation in Ne. If the WN4 in Ne is the main contributor to the fluctuation in the derived O/N2, the O/N2 ratio should peak at the Ne maximum. However, between 45°W and 45°E the WN4 fluctuation of Ne peaks at 35–30°W at 06 LT and shifts gradually to about 0–10°E before 18 LT with a diurnal maximal amplitude in the afternoon [Wan et al., 2008], which are not consistent with those of the longitudinal structures in derived O/N2 that peaks most probably at 21–33°E in the most daytime with the early diurnal maximal amplitude in the morning (Figures 2c and 2d). Actually, the radiative recombination emission contributes a small fraction to the height integration of 135.6nm emissions [cf. Strickland et al., 2004; Christensen et al., 2003]. Thus, the derived O/N2 fluctuation should not be a direct measurement of the WN4 in Ne.
 One way to induce a tidal O/N2 variation in the upper thermosphere is the fluctuation in temperature. In an equilibrium thermosphere, individual compositions distribute in static diffusion according to their own scale heights. At fixed altitudes, heavy compositions with smaller scale heights are affected more by temperature than light compositions. Then, thermal expansion would reduce the column abundance ratio of a light composition to a weighty composition for fixed altitudes. However, on a fixed pressure level the thermal expansion cannot change the O/N2 ratio [Rishbeth and Müller-Wodarg, 1999]. GUVI O/N2 is referenced to a fixed column density of N2 (close to a fixed pressure level rather than a fixed altitude) and consequently is not sensitive to thermal changes combined with static diffusion [Strickland et al., 2004]. The dynamical process provides another possible source. The WN4 fluctuation is also observed in thermospheric zonal wind [e.g., Talaat and Lieberman, 2010]. The uneven zonal winds could imply a divergence or convergence of the horizontal winds, which would be accompanied by upward or downward winds. Although there is no direct observation, the DE3 tidal vertical wind was predicted by Oberheide and Forbes [2008b] and Oberheide et al. , using Hough Mode Extension analysis of horizontal wind and temperature measurements from the TIMED and CHAMP. The O/N2 ratio could be decreased (increased) by upwelling (downwelling) motion [Rishbeth and Müller-Wodarg, 1999; Lei et al., 2010]. Thus, more knowledge about DE3 tidal vertical wind can help to understanding the behaviors of the WN4 in O/N2.
 In conclusion, the longitudinal fluctuation of equatorial O/N2 ratio contains a considerable WN4 component. The WN4 component is strong during boreal summer and autumn. In boreal winter, the WN4 component declines, accompanied with an enhanced WN3 component. The WN4 component exhibits a clear F10.7 dependence that it is generally stronger at lower F10.7 level. The seasonal variation and F10.7 dependence are consistent with those of the DE3 tide mode, indicating a potential connection with the DE3 tide mode. However, the WN4 fluctuation in O/N2 maximizes in the morning sector and its phase is almost stationary during the daytime. To satisfy the observations, we speculate that the daytime WN4 structure in O/N2 is persistence from a pattern excited at an earlier time. Some physical mechanisms were examined to explain this structure, but all mechanisms had issues. Because the ionization density in the F-region is sensitive to the variations in O/N2 ratio, the longitudinal WN4 structure in O/N2 could contribute to the longitudinal distribution of the ionospheric plasma.
 The authors thank the referees for their comments to improve greatly the manuscript. M. He thanks Z. Ren for his useful discussions. This research was supported by National Natural Science Foundation of China (40725014) and National Important Basic Research Project (2006CB806306) as well as Project supported by the Specialized Research Fund for State Key Laboratories. This study made use of the GUVI data from the Johns Hopkins University Applied Physics Laboratory. We thank PI Dr. Andrew Christensen for suggestions in the usage of GUVI data.