Electromagnetic waves and bursty electron acceleration: Implications from Freja

. Dispersive Alfv_n wave activity is identified in four dayside auroral oval events measured by the Freja satellite. The events are characterised by ion injection, bursty electron precipitation below about 1 keV, transverse ion heating and broadband extremely low frequency (ELF) emissions below the lower hybrid cutoff frequency (a few kHz). The broadband emissions are observed to become more electrostatic towards higher frequencies. Large-scale density de-pletions/cavities, as determined by the Langmuir probe measurements, and strong electrostatic emissions are often observed simultaneously. A correlation study has been carried out between the E- and B-field fluctuations below 64 Hz (the dc instrument's upper threshold) and the characteristics of the precipitating electrons. This study revealed that the energisa-tion of electrons is indeed related to the broadband ELF emissions and that the electrostatic component plays a predomi-nant role during very active magnetospheric conditions. Fur-thermore, the effect of the ELF electromagnetic emissions on the larger scale field-aligned current systems has been investigated, and it is found that such an effect cannot be detected. Instead, the Alfv6nic activity creates a local region of field-aligned currents. It is suggested that dispersive Alfv6n waves set up these local field-aligned current regions and in turn trigger more electrostatic emissions during certain conditions. In these regions ions are transversely heated, and large-scale density depletions/cavities may be

in terms of a DAW model. In sections one and two the characteristics of four Freja passages through the dayside auroral region are presented and discussed. In section three a detailed study of the ELF emissions below 64 I-Iz is presented and discussed in terms of a DAW model. In section four the expected field-aligned electric fields and propagation angles are derived from the electric and magnetic field measurements under the assumption that they are produce by dispersive Alfv6n waves. In addition, the DAW associated fieldaligned currents are investigated with regard to the effect on field-aligned current systems. A correlation study between the electric and magnetic ELF fields and field-aligned electron burst characteristics is described, and finally we make some concluding remarks regarding the overall interpretation of the observed processes in terms of wave-particle interactions caused by dispersive Alfv6n waves.

The Freja instrumentation and orbit characteristics
The Swedish/German satellite Freja passes the northern auroral region at an altitude of about 1700 km and at an inclination of 63°. It is a sun-pointing spacecraft with a spin period of 6 seconds. Detailed information about the Freja experiments can be found in a special Freja issue of Space Science Reviews (see Bochm et al., 1994;Eliasson et al., 1994;Holback et al., 1994;Marldund et al., 1994). Relevant experimental information for this study is discussed where necessary.

General characteristics of the four Freja auroral oval crossings
Data are presented from four Freja satellite passages through the dayside auroral region during ion injection events. For convenience these orbits are referred to as I, II, III, and IV, respectively, corresponding to actual orbits 5200 (November 3, 1993), 5265 (November 8, 1993), 5292 (November 10, 1993), and 6653 (February 21, 1994). A short description of the general observational characteristics is given here, and more detailed information regarding these particular data can be found in (Andersson et al., 2000). Figure 1 displays the overview of particle and wave data from all four orbits, and Figure 2 shows the corresponding energy densities as well as the de electric and magnetic field waveforms. The three first orbits occurred during moderately active conditions (Kp~2-3), while orbit IV occurred during very active conditions (Kp~6).

Ion injection and heating
All four orbits are characterised by clear ion injections with energies from 500 eV up to the TICS instrument threshold of 4.3 keV (Figure 1, panel b). The downward proton energy flux ( Figure 2, panel a) is calculated by integrating over all TICS energies from 50 eV to 4.3 kcV and over pitch angles between 0°and 90°. The time resolution of this calculated waves and bursty electron acceleration ion energy flux is 400 ms. Occasionally these fluxes are somewhat spin modulated due to limitations in pitch-angle coverage. Downward ion energy fluxes of 0.01-1 mW/m _ exist during the observed ion injections. The ion injections are accompanied by bursty electron precipitation, transverse ion heating as well as broadband ELF emissions up to the lower hybrid cutoff frequency (Figure 1). Large-scale density depletions, as measured by the Langmuir probe, are only detected during the most active orbit,/V ( Figure 1, panel 6). Intense transverse ion heating (mainly O+) is only encountered in orbit/V, for which it reaches energies of up to ,,,400 eV, while the ion heating during the other three orbits reaches at most ,,,50 eV.

Electron precipitation characteristics
The electron bursts have downward energy fluxes (calculated from pitch angles 0°-90°) of the order of 0.01-I0 mW/m 2 ( Figure 2, panels b and c), with the most intense events encountered in orbit IV. Panel b contains TESP data in the lower energy range, 12-50 eV, which corresponds roughly to energies below the local Alfv6n velocity (va). Panel c ineludes almost the full energy range of the TESP instrument (36 eV-25 keV). The time resolution of the electron measurements is 62.5 ms.
The most field-aligned electron precipitation can also be found in orbit IV (Figure 2, panel d; red line denotes 0°pitch angle, green 45°and black 90°). In all four orbits the flux in 45°and 90°pitch angle is equal but during the bursts of electrons the 0°pitch angle fluxes are much higher. In orbits II and III bursts appear more sporadically, while in orbit 13/"the electron bursts occur more frequently in time.

Coldplasma characteristics
A set of spherical Langmuir probes samples continuously the current from the plasma at a fixed positive bias voltage. This current is proportional to n_/va'_, where n_ and Te are the electron density and temperature. Usually the variation of the electron temperature is small and slow and the Langmuir probe current variation is assumed to be due solely to plasma density variations. When possible, we have used the narrowband HF Langmuir emissions near the plasma frequency, t&e, as well as the intermittent Langmuir probe bias voltage sweeps to confirm that the sampled probe current is mainly due to the variations in he, even though a variation in Te may account for a current decrease by up to a factor of 2 during active periods.
The regularly measured probe sweeps are seen as shaded grey areas in Figure 2 (panel e). In Table 1 measurements from the analysed probe sweeps, such as the plasma temperatures (Ti, Te) and the temperature ratios (TjTe), are presented together with some useful calculated plasma characteristics. As mentioned above, the ion heating is weak during orbits I/" and III, somewhat stronger in orbit I and intense in orbit IV. It should be noted that it is the whole bulk ion population that is heated, not just the energetic tail of the ion

distribution.
During orbit/V, several density drops (Figure 2, panel e) are observed simultaneously with ion injection (,panel a) and high electron count rates in the field-aligned direction as compared to count rates at other pitch angles (panel d). The three largest density drops (hereafter referred to as cavities) are centred around +55, +90 and +135 seconds into the plot. The highest electron count rates (120-130 seconds) have their main contribution from the lowest energies, below 50 eV. The electric field fluctuations (panels f and g) during orbit IV have extreme amplitudes (>3 V/m!), especiatly within the cavity regions.

Electromagnetic fields and wave characteristics
It is clearly seen from Figure 1 (panels e and 0 that the broadband ELF emissions can reach frequencies above 1 kHz and data show a large gradient, the electron distribution is isotropic with a loss cone, there is no density gradient, and the dc electric field shows just a spin modulation about zero. On orbit 111 downward bursts of electrons are correlated with sharp changes of the magnetic field.

Detailed study of the ELF emissions below 6Hz
The time series of the dc electric and magnetic field measurements were analysed by an overlapping sliding window Fast Fourier Transform (FFF) analysis ( Figure 3, panels a-e). In order to cover all frequencies below 64 Hz, four different time windows were selected (15, 6, 1, and 0.3 s corresponding to lowest non-de frequencies of 0.133, 0.333, 2 and 7.1 Hz respectively). Each time window was weighted with a Hamming filter. The longer time windows are then used at in some cases up to 6 kHz, which is close to the lower hy-_ the lower frequencies, while the shorter time windows are brid cutoff. It can also be seen that the emissions become increasingly electrostatic towards higher frequencies, while the magnetic component becomes dominant at the lowest frequencies (see panels f and g, for ac magnetic field measurements). See Wahhnd et aL (1998) for further details regarding the broadband ELF emissions. This study concentrates more on the lower frequency, more electromagnetic part, of the ELF spectrum and its relationship to electron bursts, dayside ion injections and density depletions.
The de electric and magnetic field instruments were operated somewhat differently during the four selected orbits. The dc electric field measurements normally make use of two orthogonally positioned spherical probe pairs on 10.5 m long wire booms in the spin plane (i.e. 21 m between spheres). However, during orbit IV one of the probe pairs was operated in a current collection mode allowing for 5n/n interferometry measurements, and as a consequence only one probe pair measures the dc electric field on this orbit. This explains why the electric field on orbit IV (Figure 2, panels f and g) is spin modulated. The sampling rates for orbits 1, [I and Ill are 768 samples/s for the de electric field, and 128 samples/s for the de magnetic field. On orbit IV the de electric field was sampled at 1536 samples/s and the de magnetic field at 256 samples/s.
The dc electric and magnetic field measurements have been de-spun and are presented in a coordinate system for which the x-axis is along a model magnetic field line, positive y points towards west, and positive z points towards the equator. The Freja satellite measures the electric field in the spin plane of the spacecraft and therefore only gives a 2-dimensional estimate of the total electric field. We therefore assume that the magnetic field-aligned component (Ez) is zero.
The assumption should be good since the field-aligned electric field component of a dispersive Alfv6n wave is theoretically much smaller than the perpendicular electric field component.
The field fluctuations ( Figure 2, panels • to f) are of extremely large amplitude during orbit IV, When the downward ion energy flux is close to 1 mW/m 2 the magnetic field used at the higher frequencies in Figure 3 (panels a-e). The wave spectrograms presented in Figure 3 have the same apparent time resolution as the electron data (TESP, 62.5 ms). This analysis corresponds very much to a wavelet analysis, although it is FFT-based.

Poynting flux
The wave energy per unit time and area is represented by the spectral Poynting flux (S(f) = l/poE(f) x n(f)) in Figure 3 (panel f). The Poynting flux along the Earth's magnetic field is calculated from the wave data represented in panels a through e by

S, = R_[_(E_.
Bz for each frequency component. It should be noted that the direction and magnitude of the total Poynting flux when integrated over frequency will depend on the frequency window used and their significance is unclear when treating broadband plasma wave emissions with several possible non-linear wave modes involved. We do not present the direction of the spectral Poynting flux because it shows very complex small scale variations, more suitable for detailed event studies on the time scale of seconds, and is therefore left for a future paper. Here we just note that the spectral Poynting flux is highest at the lowest frequencies, and correlates well with the downward electron energy flux (Figure 2, panels b-d).

4.2
The Alfv_n velocity and the 6E±/6B±-ratio Panels g and h in Figure 3 display as function of time the ratio between the electric and magnetic field fluctuations (SE±/6B±) and the local Alfv_n velocity (VA). The local Alfv6n velocity (VA = Bo/_ is calculated for four different ion compositions (using a H + and O+mixture with 100%, 88%, 47%, 0% H + composition) and is based on the plasma density estimated from the Langmuir probe measurements.

OrbitI
Orbit!1 OrbitIII  At the lowest displayed frequencies the ratio 6Ej_/6Bj_ is lower than VA during the three first moderately active or-

bits (I, II and III).
This suggests that dc-like field-aligned The ELF emissions are therefore more electrostatic than expected for a classical Alfv6n wave, and inertial and/or kinetic effects are probably important.
The large values at the lowest frequencies below 0.5 Hz may just be an effect of that only one probe pair was used for electric field measurements.

Dispersive Alfv6a Waves (DAW)
The term DAW is a general description covering inertial

Inertial term and kx
The perpendicular wave vector (k±) can be estimated from the Alfv6n speed (va) and the ratio 6E±/6Bx (equation 4).
From Table I While k± is independent of frequency, except for frequencies above about 40 Hz during active parts of orbit IV, kit will depend on frequency throughout the DAW dispersion relation.
The parallel wavelengths as estimated in this way are in the range of 100-1000 kin.
The parallel group velocity is close to the Alfvtn speed during orbits I, I I and I I I, while during orbit IV the parallel group velocity becomes a fraction of the Alfvtn speed, which is simply a consequence of the dispersion relation and the larger inertial terms. The perpendicular group velocity remains mosdy below 1 kin/s, at least at the lower frequencies.

Parallel electric fields
The parallel electric field associated with a DAW can be calculated in various ways. We shall adopt three methods and compare the results; one method makes use of the magnetic field measurements, the other two are calculated based on the perpendicular electric field and the above estimated wave vectors.
The first method (Figure 4, panel a)

Effect on DAW induced field-aligned current systems
The total field-aligned current variation is calculated from the fluxgate magnetometer data according to equation 6 and displayed for orbit IV in Figure 5

Particle spectral fits
The electron spectral shape is approximated by an exponentied fit to the TESP dam and characterised by the two param- where a represents the pitch angle and fo is the mea,qm_ count rate. The electron spectral fits were carried out for pitch angles 0°, 45°and 90°for which o_ was set to 0, 0.1 and 0.2, respectively, for simplicity. The parameter/3 represents the degree to which the electron population was isotropic, cies correspond to large times while the bursts occur during much shorter times.

Panel set A: field-aligned count dependence
The results of the parameter fits for an electron energy of 55 eV are compared with the electromagnetic field p_rs ( Figure 6, panel set A). Contour lines of constant density of points are drawn in the figure in order to resolve better the statistical shape of the correlation. Both the electric and magnetic field ampfitudes, and consequently the 6E±/SB±-ratio, correlate positively with the electron flux counts (parameter 7), except for the higher frequency channels of the magnetic field fluctuations. From this it follows that the Poynting flux is also correlated with the electron flux (not shown). The decrease in correlation between the field fluctuations at higher frequencies is due to the fact that the fluxgate magnetometer reaches its noise detection level, and the result is thus an ardfacL 6.3 Panel set B: electron spectral shape dependence Electron spectral shape values (l l) of less than 6 should be regardedas representing isotropic electron spectra, while > 6 indicate more field-aligned populations. The more negative the value; the more field-aligned the spectral shape, Figure 6. The more field-aligned electron populations appear when the electric and magnetic field amplitudes increase. The result is valid for all displayed frequencies.
The electron spectra at the low energy of 55 eV show a better correlation withthe ampfitude ofthewavescomparedtoa higher electron energy(not shown). Thisisnotsosurprising since the low energy field-aligned electrons are suggested to be accelerated more easily by the Mff(m waves than higher energy electrons (Andersson et al., 2000). Also, the high energy electrons, which are observed with a broad pitch angle distribution 0ow value of _, are suggested to arrive before the Alfv_n waves themselves.

Panel set C: Electron energy flux dependence
The total downward energy flux within the energy range between ,,,20-50 eV is correlated with the field characteristics. The result from this correlation follows in much the same way as the correlation with the field-aligned electron count rate at 55 eV, but is better. The contour curves indicate a good correlation between both the electric and magnetic field amplitudes with increased downward energy flux.

Panel set D: Plasma density dependence
The correlations of the plasma density with the field amplitudes show that high plasma densities (i.e. outside cavities) are not well correlated with the electromagnetic wave fields. ]Fill. 6. Correlations between various characteristics of the electrons and electromagnetic field fluctuations. Four types of correlations are presented; A-B correlate the field-aligned electron count rate at 55 eV and the electron spectral shape; C correlate the total downw_d electron am'gy flux below 50 eV; D correlates the plasma density with the field fluctuations. The field characteristics are represented by the electric, magnetic and 6E±pIB.L ratio at 5 selected frequencies (1, 2, 5, 14, and 46 Hz).
ampfitudes become large. The magnetic field strength hardly correlates with the plasma density at all, indicating that the magnetic fluctuations are not important for the formation of plasma cavities and that it is the more electrostatic component of the ELF wave activity that is related to the cavity formarion. Again, since the magnetic fluctuations are below the detection level of the fluxgate magnetometer at the highest frequency channel, the correlation loses its physical meaning at that frequency.
The good correlation between wave activity and low energy field-aiigned electron precipitation, and the anti-correlation between wave activity and plasma density, has previously also been reported by Chaston et ai. 0999).

Summary and conclusions
In the four analysed events, wave-particle interactions occurred together with intense downward proton injection. The ion injection can therefore be assumed to be produced by a common process which also creates the Alfvdn wave activity, which in turn interacts with electrons through its dispersive induced field-aligned electric field in its wave front. Altemativley an ion injection gives rise to the Alfv6n wave activity which can accelerate electrons. A scenario can be visualised in which the ratio T_/T_ plays an important role by becoming large so that the Alfv_ wave approaches the electrostatic limit and begins to accelerate more and more electrons and at the same time transversely heats the bulk ion population and raises the Ti/Te ratio further (Clark and Seyler, 1999). This positive feedback process then increases the effect of the inertial and kinetic terms further, and more energy dissipation can occur and so on.
The Freja mission was in operation during declining solar activity condition. The presence of Alfv_n waves is identified when the 5E±/6B± ratio becomes equal to or exceeds the local Alfv_n velocity. The variations of the plasma density, the Earth's magnetic field strength and the ion composition with geocentric distance cause the Alfv_n speed to have a maximum below 2 R_ geocentric distance.
During orbit IV (6653), which is an extreme dayside event recorded by the Freja satellite, the largest plasma tempera= tures were encountered and density depletions/cavities were found. Also during this orbit, the 6E±/6B±-ratio became significantly larger than the local Alfv6n speed (VA) at frequencies above the local oxygen gyro frequency (about 30 Hz). Furthermore, the largest perpendicular wave vector was found at these larger frequencies, and the wave characteristics became more electrostatic with increasing frequency. The data were therefore consistent with an interpretation in terms of dispersive Alfv6n waves Lysak and Lotko (1996), which attain slow ion acoustic characteristics at larger frequencies (e.g. Seyler et al., 1995; Wahlund et ai., 1998).
The data presented in this paper support an earlier analysis (Chaston et al., 1999) that strong ELF wave activity, especially in the electrostatic limit, is associated with larger scale plasma density depletions/cavities and at the same time enhances the low-energy electron fluxes along the geomagnetic field lines. The correlation study presented also gives a good correlation between electromagnetic field fluctuations and the presence of accelerated electrons with energies corresponding to values below the local Alfv_n velocity. Intense electrostatic ELF emissions were found to correlate well with the occurrence of density cavities.
During periods of large plasma densities, broadband ELF wave activity and bursty electrons, the integrated field-aligned net currents were zero or very small. When larger scale density depletions/cavities were detected from the Langmuir probe measurements, the integrated field-aligned net currents became more significant although remained rather small. The fluctuating field-aiigned current densities, however, are large, reaching several 100/_V/m. The largest integrated fieldaligned net currents weredetected in association with periods of low or zero ELF wave activity. At those times the fluctuations in the field-aligned current densities were on the contrary very low. A possible explanation of this behaviour is that themore isotropic electron energy distribution occurring during those times carries the current, not really afecfing the local plasma, while the large pmldlel elec_c fields associated with dispersive Alfv_ wave activity interact heavily with the local plasma, which in turn compensates for the fluctuating field-aligned currents induced by the impulsive Alfv_n waves. The flux tube associated with a large net current might be closed in the ionosphere, while the dispersive Alfv_n induced currents may have a more complexfate.