An overview of the MAVEN orbit analyzed in this study is shown in Figure 1. Panels A–H show time series plasma observations made by MAVEN centered on periapsis (the solid orange vertical line). The bar above panel A marks the plasma regions sampled by MAVEN and is discussed below. The MAVEN orbit track is shown in various planes of the MSO coordinate system in panels O1-O4. MAVEN’s orbit track is shown as the rainbow colored line in each panel, where the color maps to the rainbow colorbar beneath panel H. The whiskers in panels O1-O4 show the magnetic field vector in each plane, with the length of the whiskers proportional to the magnitude of the magnetic field. Light gray whiskers mark when MAVEN was behind the planet in each projection. The purple conic section in O2-O4 marks the statistical location of the bow shock as derived by Vignes et al. (2000).

The time series observations demonstrate that the plasma environment was highly disturbed throughout this time period due to the impact of the ICME 12–13 h earlier at about 19:25 UTC on the previous day. At the earliest times shown in Figure 1, prior to about 07:45 UTC, MAVEN sampled the undisturbed solar wind, marked by relatively little variation in the Interplanetary Magnetic Field (IMF) amplitude and vector (panels A and B), and steady solar wind velocity and density (panels D and E). The cone angle, the angle between the IMF and Mars-Sun line, was about 20° (panel C) suggesting quasi-radial IMF conditions. Despite this quasi-radial IMF configuration the standard plasma boundaries and regions appear intact, contrary to other studies of the Mars system during radial IMF conditions (e.g., Fowler C. M. et al., 2022; Chaffin et al., 2022; Zhang et al., 2024). The cause of this discrepancy is unknown and is a topic for future work. The plasma boundaries are however compressed during this time period: in particular the bow shock, which was observed at around 07:45 (denoted by the increase in magnetic field amplitude and plasma density, panels A and E), lies well below the altitude that it is typically observed at, as shown in panels O2-O4. Hanley et al. (2024) showed that during this event the highly compressed magnetosheath drove enhanced rates of electron impact ionization in the dayside ionosphere, leading to atypically high densities of C⁺ and O⁺⁺.

Downstream of the bow shock, conditions within the dayside magnetosheath were highly variable and dynamic as a result of the ICME impact: magnetic fluctuations had amplitudes of over 150 nT (panels A and B), and the background plasma density reached atypically high values of over 100 cm^-3 (panels D and E). The planetary ionosphere was encountered by MAVEN at around 08:15, observed as the increase in thermal electron density (F) and transition from the shocked magnetosheath plasma dominated by protons, to the cold thermal plasma dominated by O₂ ⁺ (G, H). Evidence of a dynamic pressure pulse caught in action compressing the ionosphere was observed just prior to this transition (marked by the red bar segment above panel A), and is described in detail in Fowler et al. (2026).

MAVEN sampled the dayside ionosphere traveling day to night; periapsis occurred at an altitude of 185 km at the terminator, at a solar zenith angle (SZA) of approximately 90°. Large amplitude magnetic structures (amplitudes ∼ 50 nT) were observed throughout periapsis and are visible centered around periapsis in the figure. These structures were driven by dynamic pressure pulses in the upstream region impacting the upper ionosphere, with the subsequent pressure pulses/magnetic structures propagating downward through the atmosphere. These structures were investigated in detail by Fowler et al. (2026), who noted that whistler waves were associated with them, which are, themselves, the focus of this study. MAVEN exited the ionosphere on the nightside of the planet, sampling a disturbed flank magnetosheath.

A zoom in between the two blue dotted vertical lines in Figure 1 is shown in Figure 2. The time range encloses five of the large amplitude magnetic structures observed at periapsis, labeled by the vertical solid black lines 1–5 (the dotted vertical line labelled ‘1b’ is discussed below). These structures are characterized by steepened leading edges that had amplitudes on the order of 50 nT, which was about 50% of the background magnetic field strength (panel A). This background field strength was atypically large during this disturbed time period: the spacecraft was located at equatorial latitudes away from the strongest crustal magnetic fields (Acuña et al., 1999), confirmed by the blue line in panel A that shows the modeled crustal magnetic field contribution from Morschhauser et al. (2014). As described in detail in Fowler et al. (2026), these structures were compressive in nature and were thus assumed to be magnetosonic, propagating at approximately the Alfvén velocity, 2–4 km s^-1 during the time range covered by Figure 2. The magnetic field was oriented horizontally with respect to the planet’s surface, consistent with the draped nature of the magnetic field about the dayside and terminator region at Mars (e.g., Crider et al., 2004). The structures, which were assumed to propagate perpendicular to the local magnetic field due to their magnetosonic nature, thus propagated vertically downward, washing over the spacecraft, which traveled at 4 km s^-1 horizontally at periapsis.

While the large amplitude structures were produced by enhancements primarily in the Bx MSO component (panel B), higher frequency and smaller amplitude oscillations were observed in all three components of the magnetic field, particularly surrounding structure one and ahead of the leading edge for structure three (panel C). The time series data shown in panel C were calculated by high pass filtering each component of the time series magnetic field observations with a 60 s sliding window. This was achieved by subtracting the average magnetic field amplitude across each window, from the value of magnetic field amplitude at the mid-point of each window. This sliding window length was chosen to encompass approximately a single structure in time. We tested window lengths of 10, 20, 30 and 60 s; our results and interpretations do not vary with window length. Zoom-ins of these higher frequency oscillations are shown in the inserts at the top of the figure, IN1 and IN2. The inserts span 25 s of observations, from between the two sets of dashed blue vertical lines in the main figure. IN1 spans the region ahead of the leading edge of structure 3, and all three components of the magnetic field show ‘clean’ sinusoidal oscillations that as we shall demonstrate are whistler waves. IN2 spans structure four; these oscillations are less clearly sinusoidal in nature and as we shall demonstrate are not indicative of whistler waves.

The data in panel C were rotated into the frame of the mean magnetic field using the same 60 s sliding window; the resulting orthogonal vector consists of one component parallel to the magnetic field (‘B_para’) and two perpendicular to it (‘B_perp1’ and ‘B_perp2’). Wavelet transforms were computed for each of these rotated time series data and are shown in panels D–F. The distinct, time dispersed nature of the whistler wave mode is observed in all three components, coincident with the higher frequency fluctuations (panel C), observed centered on structure one, and ahead of the leading edge of structure three. These fluctuations were centered at about 1 Hz and spanned frequencies from ∼ 0.1 Hz up to a few Hz. The wavelet transforms were performed over a wider time period than shown here to ensure that cone of influence effects were not present. Greater wave power was observed in the parallel component, and the two perpendicular components were not equal. The implications of these characteristics are discussed in combination with Minimum Variance Analysis, in Section 3.3.

The remaining panels G–J show the background plasma conditions. The suprathermal electron energy spectrogram is shown in panel G, with panel H showing the corresponding ratio of parallel to perpendicular temperature, which is slightly below one demonstrating that the suprathermal electron distribution is slightly anisotropic in the parallel direction relative to the local magnetic field. The thermal electron and ion densities are shown in panel I. Fowler et al. (2026) demonstrated that the troughs in ion density observed at the leading edges of each magnetic structure were associated with the squeezing of ionospheric plasma tailward as the structures propagated downward through the ionosphere. For this study, the plasma densities provided information on the ambient background conditions.

Minimum Variance Analysis (MVA) of 3D magnetic field time series observations aid in the identification of wave modes that are present. The method described by Sonnerup (1998) was applied to the MAVEN magnetic field measurements shown in Figure 2B to confirm the presence of whistler waves at magnetic structures one and three, and their absence at two, four and five. The MVA was performed over a 2 s window that captured about three oscillations of the whistler waves. Here we show examples calculated from between the dotted vertical lines in Figure 2 IN1 and IN2: the results are shown in Figure 3 (with whistler waves) and Figure 4 (without whistler waves).

Both Figures are arranged the same: panels A–C show the minimum, intermediate and maximum variance directions (B₁, B₂ and B₃ respectively), and panel D shows the magnetic field amplitude, all as functions of time. Panels E and F show hodograms of the various MVA components, where the cross and plus symbols mark the start and end of the time series respectively. The corresponding eigenvalues (and ratios of specific pairs) are shown beneath panel F and demonstrate reliable results.

Figure 3F shows three, 360° rotations that were right-hand quasi-circularly polarized, between the intermediate and maximum components. There was a gradual increase in B₃ across the time range which offset the three rotations. For each of these rotations, panel E shows an almost vertical elliptical trace between the intermediate and minimum components.

These characteristics, in combination with the wavelet analysis shown in Figure 2, are consistent with the observation of obliquely propagating whistler waves observed in the plasma rest frame. The wavelet analysis demonstrated greater wave power in the parallel direction, which implies the oblique component of propagation. The wave power in the perpendicular wavelet transforms was not equal, suggesting at best quasi-circular polarization. These interpretations are consistent with the hodograms presented in Figure 3. In addition, wave polarization analysis using the techniques described in Chaston et al. (1998) were used to derive the wave normal angle of propagation for the observed whistlers. These ranged from about 60° to 90°, indicating that these were highly oblique whistler waves. The wave polarization analysis results are not shown here as they are complimentary analysis and consistent with the wavelet transforms and Minimum Variance Analysis.

For Figure 4, Panel F does not show any clear rotation, and panel E does not show any clearly elliptical traces. This lack of features is consistent with the absence of whistler waves.

Because the whistler waves are observed centered around or at the leading edges of magnetic structures one and three, it is reasonable to assume that their generation is related to these same structures. The structures in Figure 2 are observed at altitudes between 185 and 200 km which is within the upper exobase region. The exobase is the transition region from collision dominated photochemical processes below, to electromagnetic and transport dominated processes above (e.g., Schunk and Nagy, 2000). The dayside exobase region is typically located between about 180 and 220 km at Mars (Jakosky et al., 2017). The dominant neutral species measured by MAVEN at periapsis for this event was CO₂, observed with densities ranging from about 4 × 10⁷ cm^-3 at magnetic structures one and five (the highest altitude structures shown in Figure 2), peaking at periapsis to a density of about 2 × 10⁸ cm^-3 (not shown here). Using Tables 4.4–4.6 from Schunk and Nagy (2000), the O₂ ⁺ - CO₂ collision frequency ( ν O 2 + − C O 2 ) thus ranged from about 0.02 Hz to about 0.1 Hz (highest altitudes down to periapsis respectively). The electron—CO₂ collision frequency was 10–20 Hz.

The O₂ ⁺ gyro frequency ( Ω O 2 + ) was about 0.05 Hz throughout periapsis, relatively high due to the large background magnetic field amplitude of ∼ 100 nT. The electron gyro frequency ( Ω e ) was close to 3 kHz. Because Ω e ≫ ν e 2 + − C O 2 throughout periapsis, the electrons were always magnetized. At structure one, Ω O 2 + > ν O 2 + − C O 2 by a factor of 2-3 and the ions were quasi-magnetized. At structure three, Ω O 2 + ≤ ν O 2 + − C O 2 and the ions were unmagnetized.

The whistler waves observed here bear striking similarity to those observed by Tsurutani et al. (1989) at comet Giacobini-Zinner (GZ), which were also observed at the leading edges of steepened magnetosonic waves. They postulated several potential generation mechanisms for those whistler waves, which we summarize as two categories here. The first was wave generation via plasma instability related to the pickup of heavy cometary ions: this seems unlikely for our case as there are no heavy pickup ions at such low altitudes in the Martian ionosphere.

The second generation mechanism was made in analogy to collisionless shocks and based on work by Sonnerup and Ledley (1979) and Omidi and Winske (1988). At their leading edges collisionless shocks are characterized by relatively steep gradients in magnetic field amplitude that can drive charge separation between ions and electrons across the shock front, due to differences in the ion and electron gyro radii. This charge separation sets up an electric field across the shock front that can drive currents along the shock surface (or steepened edges in our study) via E ⃗ x B ⃗ drifts. Given the similarities with the magnetic structures observed in this study - leading edges comprised of significant gradients in magnetic field strength within a mostly collisionless environment - such a mechanism seems feasible, at least for structure 1. The gyro radii of electrons and O₂ ⁺ within the 100 nT magnetic field just upstream of the magnetic structures were about 8 m and 1.5 km respectively. We used values of electron temperature T e ∼ 700 K and O₂ ⁺ temperature T i ∼ 330 K (Figure 2). The gyro radii within the steepened edges ( ∼ 150 nT magnetic field, T e ∼ 1500 K, T i ∼ 1600 K) were 8 m and 2.1 km respectively. The uncertainties in T e and T i were about 10% during this time period and do not significantly change these values. The length scale of the structures was estimated by multiplying their group velocity (2–4 km s^-1) by the time they were observed by MAVEN, about 2 s, to give a structure length scale of 4–8 km, larger than the charged particle gyro radii. The collisionless shock analogy thus seems feasible: if the ion gyro radii were much greater than the structure length scale, this would imply that the ions could gyrate ‘through’ the structures, preventing the formation of charge separation electric fields at scales relevant to the steepened edges.

In order to sustain the observed steepened magnetic edges, a bulk electrical current is required to run along them via Ampère’s Law. While the MAVEN observations cannot directly resolve charge separation on Debye scales at periapsis, MAVEN-LPW measurements may provide proxies for this at larger scales. In particular, charge separation across the steepened edges would drive localized high frequency electrostatic fluctuations, and the MAVEN-LPW instrument appears to observe such fluctuations. Figure 5 shows a combination of time series plasma observations and the electric currents measured by the Langmuir Probe instrument at the steepened edges of each magnetic structure. Panels A–C provide context of the periapsis pass, spanning the same time range as Figure 2. Panel A shows the magnetic field amplitude with the five structures again marked by the vertical solid black lines. Panel B shows the wavelet transform of B_para, where the whistler waves are again observed at structures one and three. Panel C shows the current-voltage (IV) sweeps measured by LPW. The data from panel C are plotted in panels D–M as line plots: each column (left to right) is associated with each of the five magnetic structures, with the structure number shown next to the panel letter. The top row (D-H) shows the electron dominated current measured by LPW at positive sweep voltages; the bottom row (I-M) shows the ion dominated current measured by LPW at negative sweep voltages. Each 4 s individual IV sweep is plotted as a single line in each panel. Each line is colored based on the time at which it was measured relative to each magnetic structure: the line colors map directly to the colorbar shown under panel C. The noise floor of the instrument is about 10 nA: panel J shows a good example of this noise floor, with similar amplitude fluctuations observed in every IV sweep plotted.

Both the electron and ion dominated currents to the probes (panels D–H and I-M) are proportional to the plasma density, but the electron current to the probe is larger than the ion current due to the faster electron thermal velocity. Each individual sweep consisted of 128 measurements taken over 4 s, and variability in the measured current within a single IV sweep (i.e., within an individual line) provides evidence of higher frequency fluctuations in density, which are interpreted as the presence of electrostatic fluctuations. Panels (D, I), (E, J) and (G, L) in particular show variability above the noise floor threshold in electron and/or ion currents, at the steepened edges of their respective magnetic structures. These observations thus support the presence of electrostatic fluctuations at the steepened edges of the magnetic structures, indicating charge separation at Debye length scales.

A common source of parallel propagating whistler wave generation in space plasmas are unstable anisotropic electron distribution functions that are characterized by T p e r p T p a r a greater than some instability threshold value (where T p e r p and T p a r a are the electron temperatures perpendicular and parallel to the local magnetic field respectively) (Kennel and Petschek, 1966). Figure 2H shows that the electrons are anisotropic in the parallel direction, and the ratio T p e r p T p a r a does not change significantly when the magnetic structures are observed, compared to the background values. This instability mechanism thus seems unlikely to be the cause of the observed whistler waves here, and this is consistent with the whistler waves propagating obliquely to the local magnetic field as interpreted from the wavelet and Minimum Variance analyses.

Regardless of their specific generation mechanism, Tsurutani et al. (1989) demonstrated that the whistler waves observed at comet GZ acted to reorient and smooth out the discontinuities between the up- and downstream sides of the magnetosonic steepened edges they were associated with. The observations shown in Figure 2 support the same role in this study. In panel A, structure one can be split (as a function of time) by-eye into what appears to be two steepened edges: the vertical solid line ‘1’ marks the first (clear) steepened edge and the dotted vertical line ‘1b’ marks what appears to be the peak of a second steepened edge (followed by the gradual return to background field strength). The magnetic field strength in between appears to have been ‘smoothed out’, almost removing the leading edge at structure 1b. This smoothing appears to have happened in the presence of the observed whistler wave activity, which would require obliquely propagating whistlers, consistent with our analysis. Our interpretation is that structures 1 and 1b initially looked similar to 4 and 5 (two distinct steepened structures), but the activity of the obliquely propagating whistler waves have smoothed out the second steepened edge in particular. We note that structure 1b is not necessarily a clearly defined structure, especially compared to the other structures, and this is consistent with our interpretation. Structures 2-5 are characterized by smooth reductions in magnetic field magnitude after each steepened edge. Structure 1b appears to have the smooth reduction in magnetic field strength, but at most a ‘weak’ steepened edge. Our interpretation is that the original steepened edge has been smoothed out by the observed whistler waves. There is a small increase in electron density (Figure 2I) coincident with structure 1b, a feature also consistent with our interpretation that MAVEN observed this structure towards the end of the smoothing process (in contrast to the large density enhancements that are coincident with the steepened edges that have not undergone smoothing).

The presence or not of whistler waves at the remaining structures, and their steepened nature, can be interpreted as observing these structures at different stages of their evolution as they propagate down through the ionosphere. MAVEN observed each structure at a different altitude and SZA, and subsequently each structure was observed in a different magnetization state (i.e., ratio of ion gyro frequency to ion-neutral collision frequency, see Section 4.1). The impact of this ‘sampling bias’ on whether whistler waves should have been observed associated with a particular structure or not requires further study. In addition, the chaotic and highly disturbed nature of the magnetosphere during this time period means that it is unlikely each structure was produced under identical conditions, and that all structures had traveled the same path and through the same plasma conditions. It is thus reasonable to assume that they would evolve differently, explaining why some edges had been smoothed out (1 and 1b); edge 3 was undergoing smoothing (whistler waves are present); and edges 2, 4 and 5 had yet to generate whistler waves and undergo smoothing.

To further demonstrate the role of the observed whistler waves in the evolution of the magnetic structures, a time superposed epoch analysis was carried out on all six structures and is shown in Figure 6. Each panel shows the magnetic field amplitude associated with a structure as a function of time, with time now centered on the peak in magnetic field amplitude at the steepened edge of each structure. The top row (panels A–C) shows structures that were accompanied by whistler waves (structures 1, 1b, 3); the bottom row shows structures that were not accompanied by whistler waves (structures 2, 4, 5). The two solid vertical lines in each panel enclose the widths of each leading edge, in units of seconds as observed by MAVEN. The right hand line marks the peak in magnetic field amplitude at the steepened edge. The left hand line marks the beginning of the steepened edge. This was defined as where, moving out from the steepened edge (right to left in each panel), the background magnetic field amplitude reached a minimum (or became constant). This was determined by eye and is somewhat empirical, but the results are compelling: the widths of those edges accompanied by whistler waves were 7.6, 21 and 13.6 s. These were consistently larger than the edge widths that did not have whistler waves associated with them: 1.9, 4.6 and 4.8 s. We will note the caveat here that we are limited to only six samples and these results are compelling but may not be statistically representative. Here we have assumed that the time MAVEN observed each edge for is a proxy for the width of each edge. We have not considered geometry and MAVEN’s motion relative to the steepened structures. We acknowledge that identifying the beginning of the steepened edge for structure 1b in particular is somewhat ambiguous and this is consistent with our interpretation that MAVEN observed this structure towards the end of the smoothing process acting on it. Identification of the exact start time is not crucial for our discussion here: the key takeaway is that the edge of structure 1b is clearly wider than those structures that did not have whistler waves associated with them.

A scatter plot of the steepened edge widths is shown in Figure 7, where the structure numbers are labeled next to each cross. Those structures without whistler waves associated with them (2, 4, 5) are observed by MAVEN for less than 5 s. In contrast, structures with whistler waves associated with them (1, 1b, 3) are observed for much longer, in particular structures 1b and 3, which are observed for ∼ 21 and 14 s respectively. The results are consistent with the whistler waves acting to smooth out the steepened edges.

While the observed whistler waves appear to aid in the smoothing out of the structure steepened edges, they may also interact with the ambient plasma to drive, for example, particle heating via wave-particle interactions. To determine if the observed whistler waves could Landau damp with the ambient electrons and heat the plasma, the phase velocity of the whistler waves was compared with the thermal velocity of the ambient thermal electrons, under the assumption that efficient Landau damping would occur when these two were equal.

The phase velocity of the whistler waves was calculated via the dispersion relation for the whistler wave mode, given by Equation 1: ω = k 2 c 2 ω e ω p 2 (1)

Where ω is the angular frequency of the whistler wave mode, equal to 2 π f, with f the observed frequency of the whistler waves; k is the wavenumber; c the speed of light in a vacuum; ω e = q B m e is the electron (angular) gyro frequency; ω p = n q 2 m e e 0 is the plasma (angular) frequency; m e is the electron mass; e 0 is the permittivity of free space; q is the electron charge; B is the magnetic field amplitude; n is the total plasma density.

The phase velocity V p h of the wave is defined by Equation 2: V p h = ω k k = k c 2 ω e ω p 2 (2)

From Equation 1 we can solve for the wavenumber k (Equation 3), which can then be substituted into Equation 2 to give Equation 4: k = ω ω p 2 c 2 ω e (3) V p h = c ω 1 2 ω e 1 2 ω p (4)

At periapsis, based on Figure 2, representative values for the following parameters are used: f = 0.1 Hz and 1 Hz, B = 100 nT, n = 1 × 10⁴ cm^-3. The resulting values of V p h are just below 20 km s^-1. For the observed thermal electron temperature T e of ∼ 800 K, the electron thermal velocity v t h , e = k b T e m e is about 110 km s^-1 ( k b is the Boltzmann constant). Thus at periapsis the whistler waves travel too slowly to efficiently Landau damp with the thermal electron population, consistent with the lack of observed enhanced electron temperature associated with the whistler waves (Figure 2J).

Based on Gary and Mellott (1985), Omidi and Winske (1988), Tsurutani et al. (1989), we expect the whistler wave amplitude to decrease exponentially with distance from a steepened edge if undergoing Landau damping. This was not observed at periapsis: the decrease in amplitude was roughly linear with distance (Figure 2 INS1 and INS2), and thus consistent with our calculations above. This linear decrease in wave amplitude is expected under no Landau damping as a result of wave generation and the build up of energy at the steepened edge, which has to propagate outward.

Because V p h ∝ B ω n the whistler wave phase velocity increases with increasing altitude (because B varies within a factor of two, but n decreases by several orders of magnitude). The resulting inbound altitude profiles of V p h and v t h , e are shown in Figure 8. In the figure, panels A–D show the quantities required to calculate V p h as time series; the resulting profiles are shown in panel E. The rainbow colorbar maps between the time series panels and the altitude profile (note that it is different to the timebars shown in previous Figures). At higher altitudes, around 600 km and above, V p h is comparable to v t h , e for waves at 1 Hz, and whistler waves present at these altitudes may Landau damp with the ambient thermal electron population.

A preliminary by-eye survey showed evidence for occasional whistler wave bursts at times earlier than those shown in Figure 2, including at higher altitudes between 600 and 800 km. These whistler wave signatures were not observed frequently and tended to be isolated dispersive signatures as opposed to the train of multiple dispersive signatures observed at periapsis. The ionosphere at these higher altitudes was significantly more disturbed than at periapsis, likely due to its closer proximity to the solar wind interaction region and the highly disturbed magnetosphere. Our conclusion from this preliminary survey is that while the whistler waves may have Landau damped with the thermal electrons at these higher altitudes, more detailed study is required and is left for future work.

The whistler waves were unlikely to have interacted with the ambient plasma via cyclotron resonance because the wave frequencies (0.1–1 Hz) were significantly lower than the electron gyro frequency ( ∼ 3 kHz) and greater than the O₂ ⁺ gyro frequency (0.05 Hz).

We conclude this section with a brief discussion on bigger picture impacts from this work. In particular, the magnetic structures observed here present a new mechanism through which whistler waves can be generated and propagate within the ionosphere of an unmagnetized planet. Because whistler waves typically propagate parallel to the local magnetic field, whistler waves generated in the magnetospheres of unmagnetized planets tend to be unable to propagate into the underlying dayside ionospheres. This is due to the draped nature of the global magnetic field of the induced magnetosphere: the magnetic field drapes horizontally about the dayside of the planet, preventing direct propagation of whistler waves to low altitudes. As shown here (and discussed in detail in Fowler et al. (2026)), the steepened magnetosonic structures propagated across the draped field downward into the dayside ionosphere, generating whistler waves as they did so. These whistler waves were thus generated locally within the ionosphere, where they would propagate horizontally along the draped magnetic field line on which they were generated. Such a mechanism should be feasible at other unmagnetized bodies, for example, Venus and comets, and may explain some cases of whistler waves observed at those bodies.

While the observed whistler waves in this study did not appear to interact directly with the ambient plasma, whistler waves have the potential to heat ionospheric electrons. Electron temperature is an important quantity in planetary ionospheres: it plays a key role in photochemical reaction rates that control ionospheric composition and density (e.g., McElroy et al., 1977; Andersson et al., 2010; Fox et al., 2017). It can also enhance the escape of hot atomic O via the dissociative recombination of O₂ ⁺ (e.g., Lillis et al., 2017) and drive ion escape via the generation of ambi-polar electric fields (e.g., Collinson et al., 2015; Ergun et al., 2016; Akbari et al., 2019). Thus understanding potential sources of ionospheric electron heating is crucial to understanding ionospheric structure and dynamics.