The NRL three-dimensional ionospheric simulation code SAMI3 is used to model the onset and evolution of equatorial spread F (ESF). SAMI3 is a comprehensive ionosphere model that has been modified to self-consistently solve for the global neutral wind driven dynamo electric field as well as the gravity driven electric field associated with plasma bubbles. The latter is achieved with a high resolution longitudinal grid in the pre- to post-sunset sector (i.e., 1630 MLT–2230 MLT). Initial results from the new simulation model are presented. It is shown that ESF can be triggered by pre-sunset ionospheric density perturbations, and that an existing ESF bubble can trigger a new bubble.
 Post-sunset ionospheric irregularities in the equatorial F region were first observed by Booker and Wells  using ionosondes. This phenomenon has become known as equatorial spread F (ESF). During ESF the equatorial ionosphere becomes unstable because of a Rayleigh-Taylor-like instability: large scale (10's km) electron density ‘bubbles’ can develop and rise to high altitudes (1000 km or greater at times) [Haerendel, 1974; Ossakow, 1981; Hysell, 2000]. Understanding and modeling ESF is important because of its impact on space weather: it causes radio wave scintillation that degrades communication and navigation systems. In fact, it is the focus of of the Air Force Communications/Navigation Outage Forecast Satellite (C/NOFS) mission [de La Beaujardiere, 2004].
 Two shortcomings of existing 3D ESF models is that they are limited to a narrow wedge (e.g., ∼4°) of the post-sunset ionosphere and do not include the neutral wind driven dynamo electric field. One notable exception is the work of Eccles . Eccles  embedded a high resolution grid in the time sector 1930 MLT–2130 MLT and altitude range 200–1400 km within a global ionosphere model so that the dynamo electric field was self-consistently included. The zonal resolution was ∼13 km and the altitude resolution was ∼5 km. However, a simplistic, 2D flux-tube integrated plasma model was used that only included O+ and a single molecular mixture of NO+ and O2+.
 To overcome current modeling limitations of ESF, the comprehensive, global 3D NRL ionosphere model SAMI3 has been improved to self-consistently solve for the neutral wind driven dynamo electric field as well as the gravity driven electric field associated with plasma bubbles. The latter is achieved by incorporating a high resolution longitudinal grid in the pre- to post-sunset sector (i.e., 1630 MLT–2230 MLT). The minimum resolution in this region is Δϕ = 0.0625° (i.e., a spatial scale Δx ∼ 7 km). Initial results from the new simulation model are presented in this Letter. It is shown that ESF can be triggered by pre-sunset ionospheric density perturbations, and that an existing ESF bubble can trigger a new bubble.
 The NRL 3D ionospheric code SAMI3 [Huba et al., 2008] is used in this study. The code is based on the 2D ionospheric code SAMI2 [Huba et al., 2000] and has been modified in two ways to model ESF. First, a sun-fixed coordinate system is used so the dawn-dusk line does not rotate. This requires the inclusion of the corotation electric field. Second, a high resolution longitudinal grid is used in the pre- to post-sunset sector in order to capture the evolution of equatorial plasma bubbles (see Figure 1).
 The potential equation used in this study is
Here, dipole coordinates (q, p, ϕ) are used. The definitions of the variables are given in the Appendix.
 The potential is determined as follows. First, equation (1) is solved on a uniform, coarse grid (e.g., 90 grid cells with 4° resolution) which determines the global dynamo electric field. Second, it is solved in the high resolution region (e.g., 936 grid cells with a minimum resolution 0.0625°) with the boundary values Φ1 and Φ2 specified by the coarse grid solution (see Figure 1). Using this method eliminates the use of periodic boundary conditions so that the high resolution grid (localized in the region 1630 MLT–2230 MLT) includes both the dynamo electric field as well as the gravity-driven electric field associated with plasma bubbles. Finally, the corotation potential Φcr = (B0/0.31)92/p kV is added to the solution of equation (1). Full 3D plasma transport is done on the entire grid (both coarse and high resolution) for all ion species (H+, He+, O+, N+, N2+, NO+, and O2+).
 The geophysical parameters used in this study correspond to moderate solar activity: F10.7 = 125, F10.7A = 125, Ap = 4, and the day-of-year is 80. The neutral composition and temperature are specified using NRLMSISE00 [Picone et al., 2002], and the neutral wind is specified using HWM93 [Hedin et al., 1991]. The plasma is modeled from hemisphere to hemisphere up to ±60° magnetic latitude; the minimum altitude is 90 km and the peak altitude is ∼19,000 km at the magnetic equator. We assume a magnetic dipole field aligned with the spin axis of the earth so geographic and geomagnetic coordinates are the same. The 3D model uses a grid (nz, nf, nl) = (120,200,1023) where nz is the number grid points along each magnetic field line, nf the number in ‘altitude,’ and nl the number in longitude. The simulation is run for 5 hrs. A series of Gaussian-like perturbations in the ion density are imposed at t = 0. The ion density perturbation is 5% centered at 255°, 260°, 265°, 270°, and 275° longitude. These longitudes correspond to 1700, 1720, 1740, 1800, and 1820 MLT. The perturbations have a half-width of 1° and are centered at an altitude z = 318 km.
 In Figure 2 we show electron density contours as a function of magnetic local time (MLT) and altitude (km) at 0° longitude. The overall behavior of the electron density in Figure 2 is consistent with nominal ionospheric dynamics. The ionosphere builds up after sunrise (0600 MLT) and reaches a maximum electron density in mid-afternoon (∼1500 MLT). Subsequently the ionosphere is lifted to higher altitudes because of the pre-reversal enhancement (PRE) of the eastward electric field (peaking at ∼1800 MLT). Finally, the ionosphere descends and weakens throughout the nighttime hours (i.e., 2100 MLT–0600 MLT). The important feature in Figure 2 is the development of large scale plasma bubbles (between ∼1940–2100 MLT). The initial density perturbations are amplified by a Rayleigh-Taylor-like instability that leads to the generation of a large scale ionospheric disturbances (i.e., ESF). We point out that the bubble at ∼2130 MLT was triggered by small perturbations associated with initial conditions but that the preceding bubbles were triggered by the initial, imposed perturbations.
 In Figure 3 we show plots similar to Figure 2 but for 1800–2200 MLT at four times: 0100 UT, 0200 UT, 0300 UT, and 0400 UT. The white ticks marks denote the location of the disturbances that are initiated by the ion density perturbation at t = 0. These initial perturbations lead to large scale bubble development in the post-sunset sector. The eastward drift of the bubbles is a combination of the corotational velocity and the zonal wind induced plasma velocity. Two points of interest. First, the bubbles are tilted westward consistent with observations [McClure et al., 1977; Kil et al., 2009] and simulations [Zalesak et al., 1982; Huba et al., 2009c]. Second, the plasma bubble at time 0400 UT and 2100 MLT (denoted by the red tick mark) is not triggered by an initial density perturbation. Rather, it is triggered by downward flows associated with the plasma bubble at ∼2140 MLT. We mention that the ‘down-welling’ on the sides of ESF bubbles has been observed with airglow measurements [Mendillo et al., 1985]. Lastly, the bubbles ‘diffuse away’ after ∼2140 MLT as the grid transitions from the high resolution grid to the coarse grid.
 Finally, in Figure 4 we show plots of the electron density, vertical E × B velocity (Vp), and zonal E × B velocity (Vϕ) as a function of magnetic local time (MLT) at an altitude 405 km and time 0220 UT (same as Figure 2) at the magnetic equator. The zonal velocity Vϕ has the corotation velocity removed. Figure 4 (left) shows times 0 MLT–24 MLT and Figure 4 (right) shows times 18 MLT–22 MLT. The overall behavior of the electron density, and the velocities Vp and Vϕ agree with conventional ionospheric dynamics on the diurnal time scale (Figure 4, left). The vertical velocity Vp is positive during the morning hours (∼0600 MLT–1200 MLT), is slightly negative between 1200 MLT and 1500 MLT, becomes positive and peaks at ∼1800 MLT (i.e., the pre-reversal enhancement), and is downward throughout the evening. The zonal velocity Vϕ is generally westward during the day and eastward during the night. The dynamics of the plasma bubbles is highlighted in Figure 4 (right). There are large electron density depletions between 1900 MLT and 2100 MLT (over an order of magnitude). Interestingly, on either side of a depletion is an increase in the local plasma density. Inside the plasma bubbles are strong vertical velocities as shown in Figure 4 (middle); the peak velocity is ∼200 m/s at this time. On either side of the of the strong upflows are downflows. These downflows are the cause of the plasma enhancements on sides of the bubbles; at this altitude the bubble is below the F-peak so higher density plasma at higher altitude is transported downward leading to the enhancement. In fact, it is the plasma downflow associated with the trailing edge of the bubble in Figure 3 at ∼2140 MLT that triggers the last bubble at ∼2100 MLT. And finally, there are strong zonal shears across the plasma bubbles as seen in Figure 4 (bottom right). Thus, we find complex flow patterns associated with ESF plasma bubbles.
 The first global simulation study of equatorial spread F bubble evolution using a comprehensive, global 3D ionosphere model has been presented. The NRL ionosphere model SAMI3 has been modified to self-consistently model both the neutral wind dynamo electric field (e.g., pre-reversal enhancement) and the gravity-driven bubble electric field. This has been achieved by including a high resolution longitudinal grid in the pre- to post-sunset sector. Initial results from the new simulation model have been presented. It is shown that ESF can be triggered by pre-sunset ionospheric density perturbations, and that an existing ESF bubble can trigger a new ESF bubble.
 The upgraded version of SAMI3 represents a unique resource to investigate the physics of equatorial spread F. One vexing question concerning ESF is, what processes control the day-to-day variability of ESF? We will be able to address this question by investigating the impact of background ionospheric conditions on ESF bubble development (e.g., different neutral wind models, F10.7, PRE, etc.), as well as ‘seed’ mechanisms such as gravity waves and velocity shear. Additionally, we will ‘move’ the high-resolution grid to later local times to investigate the occurrence of post-midnight ESF that has been observed by the C/NOFS satellite [Burke et al., 2009].
 There are also improvements to be made to the current model. First, the geomagnetic field will be modified to have a tilt which will introduce longitudinal effects. Ultimately, the IGRF field (or an IGRF-like field) will be incorporated into the model for a more realistic representation of the geomagnetic field. Second, we intend to couple SAMI3 with a physics-based model of the thermosphere (i.e., TIMEGCM). And finally, a high-order flux transport algorithm (i.e., the partial donor cell method) will replace the full donor cell algorithm currently being used for cross-field transport. This will allow us to capture complex bubble evolution involving bifurcation.
 The definitions of the terms in equation (1) are as follows in dipole coordinates where RE is the radius of the earth.
Here, gp is the gravity in the p direction, Ωα = eB/mαc is the cyclotron frequency of species α, and ναn is the neutral collision frequency with the α species.
 We thank J. Krall and S. L. Ossakow for a critical reading of the manuscript. This research has been supported by the ONR and NASA.