Model studies were conducted to determine the conditions that might give rise to supersonic neutral winds in the high-latitude thermosphere and that can produce neutral streams in the polar wind and plasmasphere. For the supersonic neutral wind study, we used a time-dependent, three-dimensional, global thermosphere-ionosphere model. Simulations were conducted for different seasonal conditions, solar activity levels, cross-polar-cap potentials, and convection pattern shapes. Supersonic neutral winds were found to occur at altitudes above 200 km for cross-polar-cap potentials greater than about 150 kV, although most of the supersonic winds occurred between 300 and 600 km. The Mach number of the neutral winds was typically less than 1.5, and the supersonic winds were generated within 3 h after enhanced plasma convection. These winds were primarily located in the midnight-dawn sector of the polar cap, and their dusk-to-dawn extent was of the order of 150 km. Also, during enhanced convection, the ion outflow in the polar wind increases because of the elevated plasma temperatures, and this leads to enhanced ion-neutral charge exchange reactions, and hence, to the production of neutral streams in the polar wind (Neutral Polar Wind). Likewise, on closed magnetic field lines, the upflowing ionospheric ions can charge exchange with the background neutrals, and this produces neutral streams in the plasmasphere (Neutral Plasmasphere Wind). Coupled models of the ionosphere-polar wind-exosphere and ionosphere-plasmasphere-exosphere were used to elucidate the basic characteristics of these neutral streams. An important result is that the neutral O streams typically do not have sufficient energy to escape in both the polar wind and plasmasphere, and they rain down on the global thermosphere (Neutral Rain). The downward streaming neutral O atoms pass through the exobase at 500–600 km and provide a source of momentum and energy for the thermosphere. The results presented here correspond to a focused modeling study that was presented at the Ionospheric Effects Symposium (May 2008).
 During enhanced geomagnetic activity, which typically occurs after a southward turning of the interplanetary magnetic field (IMF), the solar wind's dynamic pressure is usually enhanced and this induces changes in the magnetospheric electric fields, particle precipitation, and current systems. As the magnetic activity intensifies, the plasma convection and particle precipitation patterns expand, electric field strengths increase, and particle precipitation becomes more intense. The associated changes in the Joule and particle heating rates lead to major changes in the plasma and neutral densities, drifts, and temperatures. The cross-polar-cap neutral winds increase due to ion-neutral drag and the neutral temperatures increase due to ion-neutral frictional heating. Also, the polar wind outflow increases because of the elevated ion and electron temperatures and this leads to enhanced ion-neutral charge exchange reactions, and consequently, to the production of neutral streams in the polar wind. Likewise, on closed magnetic field lines, the upflowing ions can charge exchange with the background neutrals and this produces neutral streams in the plasmasphere. To better understand these phenomena, we conducted numerical simulations to determine the conditions that lead to supersonic neutral winds in the high-latitude thermosphere and to determine the basic characteristics of neutral streams in the polar wind and plasmasphere. The results presented here correspond to a focused modeling study that was presented at the IES2008 Symposium (May 2008).
2. Supersonic Neutral Winds in the Thermosphere
 Dynamics Explorer 2 (DE 2) satellite measurements by Balthazor and Bailey  indicated that supersonic neutral winds frequently occur in the high-latitude thermosphere between 200 and 600 km altitude. The neutral wind measurements were taken with the Wind and Temperature Spectrometer on the DE 2 satellite. While this instrument measured three-dimensional winds, the measurements of the component along the satellite's (roughly north–south) orbital path were deemed to be unreliable. Consequently, the supersonic neutral wind data were taken solely from the horizontal (zonal) component perpendicular to the satellite's motion, and hence, these data underestimate the magnitude and spatial extent of supersonic winds. Balthazor and Bailey  found that essentially all of the supersonic events were in the range 1 < M < 2. The average duration of a supersonic flow event was 20 s which, given the speed of the spacecraft, implies an average north–south spatial extent of 140 km for the region of supersonic neutral winds. The supersonic winds were typically observed in the dawn sector of the polar cap between 75° and 85° magnetic latitude, and were most numerous in the 300–400 km altitude range.
 We conducted a theoretical modeling study to determine the geophysical conditions that give rise to supersonic winds in the thermosphere. We used a time-dependent three-dimensional model of the global thermosphere-ionosphere system for this purpose (section 2.1). This model has been used in the past to study cases in which the thermosphere was perturbed by mesoscale ionospheric features. In particular, the model has been used to study the thermospheric response to single and multiple propagating plasma patches [Ma and Schunk, 1995, 1997a, 2001], multiple sun-aligned arcs [Ma and Schunk, 1997b], equatorial plasma bubbles [Schunk and Demars, 2003, 2005], theta-aurora [Demars and Schunk, 2005], upwelling in the dayside cusp due to ion heating [Demars and Schunk, 2007a], and upward propagating waves from the lower atmosphere [Schunk et al., 2008].
2.1. Thermosphere-Ionosphere Model
 An extensive description of the thermosphere-ionosphere (T-I) model is described by Ma and Schunk . Here, we note that the model calculates a simultaneous solution of the neutral gas equations of continuity, momentum, energy, and mean molecular mass, which produces global distributions of the mass density, the temperature, and all three components of the neutral wind over the altitude range from 97 to about 500 km. The differential equations are solved in a spherical coordinate system fixed with the rotating Earth using a multidimensional flux-corrected transport (FCT) technique [Zalesak, 1979]. The model uses height as its third (vertical) degree of freedom instead of the pressure coordinate used in some previous thermosphere models [Dickinson et al., 1981; Fuller-Rowell and Rees, 1980], and nonhydrostatic equilibrium flows can be modeled. Note that the FCT method is a well known numerical technique that was specifically designed to handle subsonic, transonic, and supersonic flow, and hence, is ideally suited for this study. Typically, the model resolution is 0.5 degree in latitude and 3 degrees in longitude, and the altitude range is divided into 49 layers. However, since the problem addressed in this study does not involve small-scale ionospheric features, we used a lower resolution (3° in latitude, 5° in longitude, 49 altitude layers) to generate the results presented below.
 The magnetic field and convection electric field are treated as inputs to the thermosphere-ionosphere model. The magnetic field is calculated from a tilted dipole that rotates with the Earth [see Ma and Schunk, 1995]. The convection electric field is obtained for some cases from the Volland model [Volland, 1975] and for other cases from the Weimer model [Weimer, 2001].
 The ionosphere affects the momentum and energy balance in the thermosphere via the ion drag force and the ion-neutral frictional heating terms in the transport equations. As in our previous studies, the International Reference Ionosphere (IRI) model [Bilitza, 1990] is used to obtain global ion density distributions. In addition to the IRI background ionosphere, ionization due to auroral particle precipitation is accounted for using the method of Roble and Ridley . There is no feedback from the thermosphere model to the ionosphere. However, as noted below, the supersonic neutral winds develop in less than 3 h after the convection electric field increases, and therefore, the feedback is not expected to change our basic conclusions.
2.2. General Simulation Procedure
 The largest plasma convection velocities typically occur for southward interplanetary magnetic field (IMF). In this case, the magnetospheric electric potential pattern has two cells, which act to induce antisunward flow over the polar cap and return flow at latitudes below the auroral oval. However, the plasma is also subjected to a corotational electric field, and when this is added to the magnetospheric electric field the resulting plasma drift pattern is similar to that shown in Figure 1. As the plasma drifts across the polar cap, ion-neutral collisions act both to frictionally heat the neutrals and to accelerate the neutrals in the antisunward direction (Figure 2). Whether or not the neutral flow becomes supersonic depends on the overall thermosphere-ionosphere conditions.
 Solutions to the thermosphere-ionosphere model were obtained in three steps. First, an initial estimate of the thermosphere and ionosphere conditions corresponding to a particular geophysical case was obtained from the MSIS [Hedin, 1991] and IRI models, respectively. This required the selection of values for various physical parameters, including the date, solar F10.7 flux, sunspot number, Ap index, etc. This initial estimate was used as the starting point for the global, time-dependent thermosphere-ionosphere simulations. Next, the global thermosphere-ionosphere model was run until a diurnally reproducible pattern was established; typically 4 days. The diurnally reproducible pattern is different for different seasonal conditions and solar activity levels. For all geophysical cases considered, the diurnally reproducible pattern was obtained using the Volland  convection model and assuming a cross-polar-cap potential of 40 kV. In the final step, the cross-polar-cap potential was increased to some value in the range 100–200 kV, and the model was run for a period of time to see if, and where, the thermosphere exhibits supersonic flow. In this final step, the convection pattern either remained a Volland pattern, but with a larger cross-polar-cap potential, or was changed to a Weimer pattern for southward Bz and either positive or negative By. After the change the pattern was held fixed. For all convection patterns that were adopted, the auroral precipitation pattern was adjusted to be consistent with the convection patterns. Also, for all cases, the effect of increasing the potential was fully seen after only 3 h. The results presented below are for 3 h into the simulations, which corresponds to 3 UT.
 The Mach number of the neutral gas is defined to be the neutral wind speed divided by the thermal speed. The thermal speed is given by the expression (kT/m)0.5, where T is the neutral temperature, m is the mean neutral mass, and k is the Boltzmann constant. At 300 km, the neutral mass is atomic oxygen.
2.3. Thermosphere-Ionosphere Simulations
 Simulations were conducted for three seasons (June solstice, December solstice, equinox), three solar activity levels (F10.7 = 70, 150, 220), three cross-polar-cap potentials (100, 150, 200 kV), and both Volland and Weimer convection patterns for southward IMF [Demars and Schunk, 2007b, 2008]. First, a simulation was conducted for a baseline case corresponding to equinox conditions, moderate solar activity (F10.7 = 150), a Volland potential pattern, and a cross-polar-cap potential of 150 kV. Subsequent simulations were then compared to this baseline case. Here, however, only three representative cases are presented and then a summary of the results is given.
Figure 3a shows snapshots of the Mach number at 3 UT in the northern polar region at 200, 300, 400, and 500 km altitude for the baseline case described above. The coordinate system is magnetic latitude versus magnetic local time (MLT), with the north magnetic pole at the center of each dial. The neutral wind is subsonic at 200 and 300 km. At 400 and 500 km, a region of weak supersonic flow occurs at about 75–90 degrees north magnetic latitude in the midnight-dawn sector. Figure 3b shows the corresponding ion velocities (left) and neutral winds (middle) at 400 km altitude. The ion velocities are larger in magnitude than the neutral winds at most locations in the polar cap. The tips of the neutral wind vectors are roughly parabolic along a given dusk-dawn line over the polar cap and this is similar to what has been observed in DE 2 data (Figure 2). Also shown in Figure 3b (right) are profiles of the Mach number along a line from dusk to dawn through the peaks of the Mach number distributions shown in Figure 3a. The profiles are at 400 (solid), 300 (dashed), and 200 (dotted) km. The profile at 500 km is omitted for clarity because it is too close to the profile for 400 km. The Mach number profile at 400 km shows a limited region of weak supersonic flow (1 < M < 1.08) that mostly appears on the dawn side of the noon-midnight meridian.
 The results shown in Figures 4a and 4b are for the baseline conditions except that the cross-polar-cap potential is increased to 200 kV. Ion velocities (Figure 4b) are considerably enhanced due to the stronger convection. This results in increased neutral winds due to ion-neutral drag in regions where the ions and neutrals move in roughly the same direction (Figure 4b). The Mach numbers are increased at all altitudes (Figures 4a and 4b), with the maximum Mach number at 400 km reaching about 1.35. The region of supersonic winds extends below 300 km altitude.
 Several simulations were conducted using Weimer electric potential patterns for Bz < 0 and, hence, for two-cell convection. For both By < 0 and By > 0, we considered different solar activity levels (F10.7 values ranging from 70 to 220), different seasonal conditions (summer, winter, and equinox), and different values for the cross-polar-cap potential (100–200 kV). Since the cross-polar-cap potential is not an input parameter of the Weimer model, it was necessary to vary the actual Weimer inputs (angle in Bz-By plane, Bt, dipole tilt angle, and the solar wind velocity and density) in order to produce the desired potential drops. A simulation for By > 0 is presented because the corresponding Weimer pattern is asymmetric with regard to the shape and orientation of the convection cells, which is in sharp contrast to the symmetric Volland patterns used in the simulations discussed above. Figure 5 (left) shows an asymmetric Weimer pattern for By > 0. The pattern corresponds to a cross-polar-cap potential of 150 kV and was obtained from the Weimer model using an angle of 135 degrees in the Bz-By plane, a Bt value of 25 nT, a tilt angle of 0 degrees, a solar wind velocity of 440 km/s, and a solar wind density of 9 cm−3. As before, the simulated auroral oval (Figure 5, right) has been adjusted in size and shape to be consistent with the potential pattern.
 A thermosphere-ionosphere simulation for summer solar maximum conditions is presented in Figures 6a and 6b for By > 0 and a cross-polar-cap potential of 150 kV. The location and curved shape of the regions of high, neutral gas Mach numbers that are shown in Figure 6a result from the asymmetry of the Weimer pattern for By > 0 and illustrate the dominant role played by ion drag in accelerating the thermosphere. The ions drift around the sunward border of the dusk convection cell and then turn back and accelerate through the curved "convection throat" between the cells (Figure 5, left). As they drift in this manner, they transfer momentum to the neutrals. This not only accelerates the neutral gas, but also imposes a westward component to its predominantly antisunward flow (Figure 6b). High Mach numbers and supersonic flow are predicted for the neutral gas at altitudes as low as 200 km (Figure 6b). In the 300–400 km altitude range, the Mach number of the neutral wind reaches a value of about 1.25 in the convection throat (Figures 6a and 6b). Both ion and neutral velocities reach magnitudes of 1 km/s or more in the region of rapid convection between the cells (Figure 6b). In general, for potential drops greater (less) than 150 kV, we observe larger (smaller) neutral Mach numbers, for both the By < 0 and By > 0 cases.
3. Neutral Streams in the Ionosphere and Plasmasphere
 In the upper atmosphere, between approximately 500 and 1500 km, charge exchange reactions are an important part of the ion-neutral chemistry. In this region, O+ and H+ ions interact with the background and geocoronal O and H with which an electron is exchanged and the ions are converted into neutral stream particles, Hs and Os (Figure 7). Charge exchange plays a vital role along both open and closed fields, at high, mid, and low latitudes. Along open field lines, ions are accelerated due to ambipolar electric fields. The ions then interact with the background neutral atoms and exchange electrons, becoming neutral atoms with a bulk motion in the general direction of the field line at the points where the charge exchange reactions occurred. If they have sufficient energy, they will populate the magnetosphere, but if not, they will return to the upper atmosphere due to gravitational attraction. Along closed field lines, the ions tend to flow from one hemisphere to the other, and this occurs for all solar cycle conditions, seasons, and levels of magnetic activity. As with the ions on open filed lines, the ions that flow along closed field lines interact with the background neutrals, undergo charge exchange reactions, and become neutrals moving in a generally field-aligned direction.
 The lighter neutrals produced in charge exchange reactions can have enough energy to escape. These lighter neutrals tend to populate the inner magnetosphere and the plasmasphere, where they can interact with the plasma in the plasmasphere. The heavier neutrals tend to be produced with lower energies, and will not have sufficient energy to escape; they fall back to earth interacting with the thermosphere. At high latitudes, a fountain like effect can be seen with charge exchange occurring in the dayside oval and cusp, where ions are accelerated up open field lines and are then acted upon by the convection electric field, which under a southward interplanetary magnetic field will generally cause an antisunward motion of the ions. The neutrals produced will have a general upward, antisunward flow, but as they move across the polar cap the heavy neutrals will fall back to the Earth, creating a neutral rain on the high-latitude thermosphere. On closed field lines, the same general phenomena occur. Ions move up field lines in one hemisphere, undergo charge exchange reactions, turn into neutrals, and then rain down on the opposite hemisphere. This rain could be a source of energy for the thermosphere, due to the fact that stream particle energies are generally higher than typical energies of the thermospheric neutrals.
 Measurements have shown the effects of high-energy neutral stream particles impacting the thermosphere. Zhang et al.  used FUV images to show that the high-energy ions in the ring current can undergo charge exchange reactions and produce what they termed neutral particle aurora at mid and low latitudes. This is the same type of phenomena we are studying, except that in our case the energies involved are much lower and happen on a continuous basis rather than only during strong magnetic storms.
3.1. Neutral Polar Wind
 Simulations with our 3-D, time-dependent model of the neutral and ion polar winds indicate that substantial neutral and ion fluxes are created during storms at altitudes above about 500 km, with the neutral fluxes larger than the ion fluxes [Schunk and Sojka, 1997; Gardner and Schunk, 2005, 2006a], and these results are consistent with measurements [Wilson et al., 2003]. As the H+ and O+ ions drift upward in response to storm heating, they are accelerated to velocities as high as 10–20 km/s for H+ and 3–5 km/s for O+. Charge exchange of the upflowing ions with the background neutral atmosphere (thermal and hot geocoronal neutrals) acts to produce energetic streaming H and O neutrals (Figures 7 and 8) . Both H+ and H have sufficient energy to escape, but O+ and O eventually reverse direction and head toward the Earth.
 The recent studies of Gardner and Schunk [2004, 2005, 2006a, 2006b, 2007] have demonstrated the importance of charge exchange reactions with regard to high-latitude outflow processes. A new study by Gardner and Schunk [2008a] used measured storm time convection and precipitation patterns to drive a 3-D neutral and ion polar wind model, which revealed a pulsing of the total vertical flow rates for all species studied. The cross-polar-cap potential and the energy input due to precipitating electrons (Figure 9) shows that the pulsing nature is in the drivers, which pulse with about a 50-min period. Only this single event has been studied to date, so it is not clear whether this pulsing is typical of magnetic storms, or whether this is a one of a kind or a rare occurrence. Another interesting feature in the input data is the apparent shift in the maxima. In Figure 9 (top), the cross-polar-cap potential has its largest values near 5:00 UT, whereas the largest values for the electron precipitation energy input occur near 7:00 UT. On top of the general signal in both cases is superimposed the approximate 50-min oscillation.
 The resulting ion and neutral ‘total’ vertical fluxes are shown in Figure 10 at 1500 km altitude. The total vertical fluxes are obtained by integrating the vertical fluxes over the entire polar region. Figure 10a shows that the vertical H+ fluxes exhibit predominantly upward fluxes throughout the entire storm period. The H+ velocities reach 10–20 km/s, thus allowing it to penetrate deep into the magnetosphere. The O+ total vertical fluxes, on the other hand, tend to oscillate more in an upward/downward motion, with the O+ not having enough energy to escape. Thus, the O+ tends to rise in the oval, and to descend over the polar cap and equatorward of the oval.
 The neutral stream particles show a similar nature to the ions. However, the Hs particles always show an upward total integrated flux and have velocities of about 10 km/s, thus allowing them to penetrate deep into the magnetosphere. Figure 10c shows the total vertical flux for the Hs particles, again showing the oscillatory nature. The total vertical flux for the neutral Os particles is shown in Figure 10d. The Os particles maintain a negative total vertical flux at 1500 km at all times. This is due both to Os falling in the polar cap and equatorward of the oval, and to the fact that charge exchange occurs above the altitude level where the fluxes were integrated.
 The general nature of the Os particles in a 2-D slice across the polar noon-midnight meridian is shown in Figure 11. The oval is located at approximately 70° on the dayside, and near 65° (115° in Figure 11) on the night side. The Os particles are created in the dayside oval, or from ions drifting across the polar cap due to convection electric fields, thus producing the antisunward flow between 70° and 95°. Since the O+ ions are not energized to high values, the neutrals produced are pulled by gravity, and hence, the Os particles fall back to Earth as they cross the polar cap. Again, a similar process occurs near the night-side oval at 115°. The Os particles are created above the oval and are insufficiently energized to continue flowing vertically; they therefore end up raining down equatorward of the oval. Most, if not all, of the Os particles produced at low altitudes (<800 km) move in the horizontal direction, and subsequently, they tend to rain down on the thermosphere. The neutral stream particles have initial velocities of 1–3 km/s, and their energy is deposited in the thermosphere.
3.2. Neutral Plasmasphere Wind
 The plasmasphere is a torus-shaped volume that surrounds the Earth and contains a low temperature, high-density plasma that has its origins in the ionosphere. The plasma in this region corotates with the Earth, but it can flow along B from one hemisphere to the other [Schunk and Nagy, 2000]. During solstice conditions, the flow is primarily from the summer to the winter hemisphere. However, during geomagnetic storms, the outer regions of the plasmasphere are peeled away, and then the ionospheric flow is upward from both hemispheres day and night as the plasmasphere refills, which takes about 10 days. Typically, storms and substorms occur frequently, and therefore, the outer plasmasphere is in a continual state of refilling. As the H+ and O+ ions flow upward along B, they can charge exchange with the background neutral atmosphere, including thermal and hot geo-coronal neutrals, and this acts to create energetic neutral streams in a manner analogous to what occurs in the neutral polar wind (Figure 8).
 A preliminary study of the neutral streams in the plasmasphere was conducted using a coupled ionosphere-plasmasphere-exosphere model. However, before we discuss the science, we will briefly discuss the status of the coupled model. The Ionosphere-Plasmasphere Model (IPM) is a global model that covers geomagnetic latitudes from about 60° N to 60° S and equatorial crossing altitudes from 90 to 20,000 km [Schunk et al., 2004; Scherliess et al., 2004]. The model includes chemical, radiation, and transport processes that are similar to those in our global Time-Dependent Ionosphere Model [Schunk, 1988; Sojka, 1989], but the IPM includes several ions. At E region altitudes, chemical equilibrium is assumed and the continuity equations for NO+, O2+, N2+, and O+ are solved simultaneously for the ion densities. At F region altitudes, the continuity and momentum equations for H+, He+, and O+ are solved along B for individual plasma flux tubes. The flux tubes are followed as they drift through a moving neutral atmosphere perpendicular to B due to corotational and dynamo electric fields. The 3-D nature of the model is obtained by following many plasma flux tubes while keeping track of their positions at all times. The IPM adopts the International Geomagnetic Reference Field, and uses the empirical model of Titheridge  for Te and Ti, and the Scherliess and Fejer  and Scherliess et al.  model for equatorial electric fields. The neutral stream model is a new high-order numerical model and for this study is currently 2-D. The continuity and momentum equations for H and O are solved as a function of altitude and latitude for a given longitude using a fifth-order WENO technique, and the ion parameters are provided by the IPM. Charge exchange reactions of H+ and O+ with both thermal and geo-coronal neutrals are taken into account. Note that the IPM grid follows dipole magnetic field lines, while the grid for the neutral streams is altitude versus latitude. Hence, the grid for the neutral streams does not match the dipole grid. In Figure 12 the top left and right-hand regions of each plot are not relevant to the physical solution.
 For our initial model study, day 90 of 1998 was selected, which corresponds to moderate solar activity and low magnetic activity. The resulting model run is shown in Figure 12. The ions flow from the southern hemisphere (left side of each plot) to the northern hemisphere (right side of each plot) as seen in Figures 12a and 12b. The ions then undergo charge exchange reactions with the background neutrals, creating streaming neutrals by this process, which then rain down on the thermosphere, similar to what happens at high latitudes. The neutral Os particles rain down on the thermosphere with velocities approaching 200 m/s. For the neutral H stream particles there is a critical region between 600 and 800 km. Above this region the neutral H streams produced in the charge exchange reactions flow upward, whereas those produce below this level flow downward. The neutral O stream particles, on the other hand, tend to follow parabolic orbits from the southern hemisphere to the northern hemisphere, thus raining back down into the thermosphere.
3.3. Neutral Rain on the Thermosphere
Figure 13 shows a schematic representation of the global flow of neutral Os atoms. Figure 13 (right) shows the high-latitude open field line region, where the ion and neutral polar winds are important. The ions (arrows with an i) flow along magnetic field lines and are convected, in this case, antisunward across the polar cap. The ions then undergo charge exchange reactions and are converted into neutral stream particles (arrows with an n). In the cusp region, toward the left, the neutrals act to produce a fountain, with some of the neutrals flowing in the antisunward direction due to convection of their parent ions, moving into the polar cap, and the rest just falling back in a symmetrical way around the cusp. In the polar cap, the ions are driven by the convection electric field, again in the antisunward direction, and charge exchange reactions produce neutrals primarily moving in the antisunward direction. In the nocturnal auroral oval, a fountain effect is produced with ions moving up the field lines on the edge of the convection zone, and the resulting streaming neutrals fall back toward the Earth like rain, forming a symmetrical pattern around the auroral zone.
Figure 13 (left) shows what occurs in the plasmasphere on closed magnetic field lines. The ions (arrows with i) flow up the closed magnetic field lines and then some of them undergo charge exchange reactions, turn into neutral stream particles, and subsequently, they rain down on the equatorial thermosphere. The energy of the produced neutrals, the time of year, and the geomagnetic activity all play a role in determining the flow of the neutrals in the plasmasphere, and therefore, the exact distribution and location of the neutral rain are expected to vary appreciably.
 A comparison of Figure 11 and Figure 13 (right) shows that the resulting neutral Os flows closely relate to the schematic; however, a distinct cusp was not included in the polar wind simulation. Figure 11 clearly shows the neutrals flowing antisunward across the polar cap and raining down on the thermosphere in the region of the pole, and the fountain effect in association with the night-side oval.
 Comparing Figure 12 with Figure 13 (left), it is apparent that the general behavior is again reproduced. In Figure 12 the ions flow from the southern hemisphere to the northern hemisphere, and so the neutral stream particles produced in charge exchange reactions also flow from the southern hemisphere to the northern hemisphere. The neutrals produce in the charge exchange reactions then rain down on the thermosphere across the entire equatorial region, where they add momentum and energy to the thermosphere as their velocities are reduced through collisions with thermospheric neutrals.
 Model studies were conducted to determine the conditions that might give rise to supersonic neutral winds in the high-latitude thermosphere and that can produce neutral streams in the polar wind and plasmasphere. For the supersonic neutral wind study, we used a time-dependent, three-dimensional, global thermosphere-ionosphere model. Simulations were conducted for three seasons (June solstice, December solstice, equinox), three solar activity levels (F10.7 = 70, 150, 220), three cross-polar-cap potentials (100, 150, 200 kV), and both Volland and Weimer convection patterns for southward interplanetary magnetic field. The simulations therefore covered a wide range of background ionospheres, thermospheres, and convection patterns. The simulations indicated that the Mach number of the neutral wind increases with increasing solar activity, increasing magnetic activity and in going from winter to summer conditions. Supersonic neutral winds were found to occur at altitudes above 200 km for cross-polar-cap potentials greater than about 150 kV, although most of the supersonic winds occurred at and above 300 km. The Mach number of the neutral winds was typically less than 1.5. The supersonic winds appeared within one-half to 3 h after the enhanced plasma convection was imposed on the thermosphere. These winds were primarily located in the midnight-dawn sector of the polar cap and their dusk-to-dawn extent was of the order of 150 km. These modeled features are in agreement with the DE-2 satellite measurements.
 Charge exchange reactions are important in the upper atmosphere on both open and closed magnetic field lines in the altitude range of from 500 to 3000 km. As ions are accelerated along magnetic field lines due to ambipolar electric fields, or across field lines due to convection electric fields, they can undergo charge-exchange reactions, exchanging electrons with neutrals in close proximity, and then becoming neutral stream particles. The neutral stream particles acquire the velocities that their parent ions had just prior to the charge exchange reactions, and hence, are at higher energies than the thermal background neutrals. At high altitudes on open field lines, the charge exchange reactions produce both light neutrals that can populate the inner magnetosphere and heavier neutrals that then rain down on the high-latitude thermosphere, depositing energy into the thermosphere system. A similar process occurs at mid and low latitudes. The ions on closed magnetic field lines are accelerated along magnetic field lines due to ambipolar electric fields during plasmaspheric refilling and across magnetic field lines due to E × B drifts. Since the ions are on closed field lines, they typically do not achieve the velocities they do on open field lines. However, the plasmasphere densities are greater than those on open field lines and the neutral stream particles produced will again rain down on the thermosphere. The downward streaming neutral Os atoms pass through the exobase at 500–600 km, and they could be an important source of momentum and energy for the thermosphere.