Investigation of Mars ionospheric response to solar energetic particle events

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi:10.1029/2012ja017671, 2012 Investigation of Mars ionospheric response to solar energetic particle events Demet Ulusen, 1 David A. Brain, 2 Janet G. Luhmann, 3 and David L. Mitchell 3 Received 29 February 2012; revised 8 October 2012; accepted 13 October 2012; published 11 December 2012. [1] We investigate the effects of solar energetic particle (SEP) events on the Martian ionosphere using observations from the Mars Global Surveyor (MGS) Electron Reflectometer (ER) and Radio Science (RS) experiments. Although MGS/ER is not designed to measure solar storm particles, it detects SEPs as increased instrument background. Using this proxy for SEP fluxes near Mars, we compare electron density profiles obtained from the RS experiment during periods of high and low SEP activity. Six case studies show no clear evidence for an increase in the ionospheric electron density between 200 and 100 km altitudes. However, 4 of the 6 events show a small increase in electron density below 100 km altitude during SEP events, suggesting that high-energy (10 20 kev) electrons may cause ionization in the lower ionosphere. We also observe an 25% decrease in the ionospheric electron density between 100 and 120 km altitude for the two strongest events, suggesting that SEPs trigger a process that increases electron loss in this altitude range of the lower ionosphere. However, we cannot be confident from only two events that this effect is caused directly or indirectly by increased SEP fluxes. A statistical study confirms the case study results, but not over all solar zenith angles. Additionally, we observe depletions in the topside ionospheric electron density at some solar zenith angles, which can be explained by compression of the ionosphere by the passing CME. Citation: Ulusen, D., D. A. Brain, J. G. Luhmann, and D. L. Mitchell (2012), Investigation of Mars ionospheric response to solar energetic particle events, J. Geophys. Res., 117,, doi:10.1029/2012ja017671. 1. Introduction [2] Sudden increases in the solar radiation and energetic particles associated with solar activity create a variety of space weather disturbances in the near space environment of planets. Space weather storms are generally associated with the solar eruptions known as Coronal Mass Ejections or CMEs [Hudson et al., 2006; Howard and Tappin, 2009, and references therein]. These ejections of plasma and magnetic field produce disturbances that may travel at up to 1500 km/s through the ambient solar wind in the inner solar system. As they propagate, they interact with the solar wind to produce a leading shock wave followed by solar wind plasma and field, the CME ejecta driver which may appear as a magnetic flux rope. The passage of these disturbances may last 1 2 days. Associated with these events are Solar Energetic Particle (SEP; having energies between 10 MeV and 1 GeV) events [Reames, 1999, and references therein]. The effects of major 1 Space Technologies Research Institute, Ankara, Turkey. 2 Laboratory for Atmospheric and Space Physics, University of Colorado Boulder, Boulder, Colorado, USA. 3 Space Sciences Laboratory, University of California, Berkeley, California, USA. Corresponding author: D. Ulusen, Space Technologies Research Institute, ODTÜ Yerleşkesi, 06531 Ankara, Turkey. (demetulusen@gmail.com) 2012. American Geophysical Union. All Rights Reserved. 0148-0227/12/2012JA017671 events encountering Earth are well studied through abundance of observations and numerical simulations [Tsurutani et al., 2005; Cully et al., 2003; Green et al., 2006]. The influence of such events on near-mars space has been investigated in a few studies, where it was found that Mars response is different from Earth s, mainly because Mars has no global magnetosphere [Crider et al., 2005; Espley et al., 2005; Mendillo et al., 2006; Haider et al., 2009; Morgan et al., 2006; Futaana et al., 2008; Leblanc et al., 2002; Falkenberg et al., 2011; Lollo et al., 2012; Sheel et al., 2012]. [3] For both Mars and Earth, different types of effects are expected for the different components of a storm. For example, at Earth the plasma and field disturbance produces a geomagnetic storm and associated aurora, radiation belt enhancements and ionospheric disturbances [e.g., Dmitriev and Yeh, 2008; Green et al., 2006], while the SEP event produces additional ionization in the polar cap and ozone destruction in the mesosphere and sometimes down to the stratosphere [e.g., Tsurutani et al., 2005; Cully et al., 2003]. [4] Previous investigations of space weather effects at Mars have examined different aspects of the space weather responses. Using MGS magnetometer observations, Crider et al. [2005] and Espley et al. [2005] investigated the effects of the Halloween 2003 solar storms, which included exceptional CMEs, on the Martian plasma environment. In their work, they suggest an increase in atmospheric escape rates during these extreme events, which results from the 1of19

compression of the solar wind plasma interaction region [Crider et al., 2005] and from an increase in magnetic field fluctuations near the ion gyrofrequencies [Espley et al., 2005]. McKenna-Lawlor [2008] applied HAFv.2 (Hakamada-Akasofu_Fry, version 2), a statistically tested numerical model that provides useful predictions of shock arrivals at Earth, to Mars and satisfactorily predicted four shock events at Mars. These studies considered mainly the regions above the ionosphere. Based on MGS radio science measurements, Mendillo et al. [2006] showed that soft X-rays originating from solar flares (within 0.1 5 nm) resulted in an ionospheric density enhancement up to 200% below 110 km altitude. Later, using similar data, Mahajan et al. [2009] observed that X-ray flares form a well defined E layer peak, which is not typically seen, and some of these flares coincide with enhancements in both E and F1 region electron densities. Moreover, Haider et al. [2009] showed that the total electron content of the Martian ionosphere is increased by factors of 6 to 10 due to soft X-rays during a particularly violent solar event. Recently, Lollo et al. [2012] has developed an ionospheric model which successfully reproduces the vertical profiles before and during a solar flare. These studies were not concerned with the additional effects that a plasma and field disturbance, or its associated SEPs, might have on the ionosphere. In particular, although these latter investigations confirmed the effect on the Martian ionosphere of increased solar photon fluxes during space weather events; the direct influence of solar energetic particles (SEPs) has not been established. [5] SEPs, composed of mainly protons and electrons accelerated in solar flares or at CME shocks, were first measured at Mars by the Phobos 2 mission [McKenna- Lawlor et al., 1992, 2005] and then by the Martian Radiation Environment Experiment (MARIE) instrument on the Mars Odyssey spacecraft [Zeitlin et al., 2004]. The MARIE instrument, designed to monitor the radiation environment of Mars, detected a number of space weather events during its operation period until it ceased functioning during the large Halloween 2003 solar storm. These SEP measurements show an orbital modulation, which is attributed to Mars shadowing or absorbing part of the SEPs in particular locations around the planet. Using other observations from the Mars Express (MEX) MARSIS experiment, Gurnett et al. [2005], Morgan et al. [2006, 2010] and Espley et al. [2004] reported the occasional ionospheric absorption of the radar signals reflected from the surface. These several-day-long absorptions suggested increased ionization in the Martian upper atmosphere and were found to be correlated with the SEP events detected at Mars by MGS ER [Espley et al., 2007]. The MARSIS blackouts, as reported by Morgan et al. [2006, 2010] and Espley et al. [2007], implied plasma density enhancements at some altitude below the main peak by some unspecified amount. Withers [2011] presented some calculations of what altitudes and enhancements are consistent with the observations. In addition, using MEX ASPERA-3 observations at Mars during the major X9.0 solar flare event in December 2006, Futaana et al. [2008] showed a correlation between the detection of SEP backgrounds and an order-ofmagnitude increase in the heavy atmospheric ion escape at Mars. In the analysis of 26 months of data from the MEX MARSIS instrument, Lillis et al. [2010] found that the total electron content (TEC) of the Martian ionosphere increases during disturbed solar and space weather conditions and also suggested that solar energetic particle events may cause shortlived increases in TEC of about 10 5 m 2 at all SZAs. [6] In addition to these observations, several numerical models have been developed to explore both the nature of the SEP event observations at Mars and their potential effects. Leblanc et al. [2002] investigated the effects of a typical SEP event on the Martian atmosphere using test particle simulations. In this work, SEPs are predicted to penetrate the bow shock and the magnetic barrier from the piled-up interplanetary field on the dayside. Based on a prototypical SEP event energy spectrum and fluxes, they estimate that the energy deposition of a typical SEP event in the atmosphere is an order of magnitude less than the solar EUV (extreme ultraviolet) energy deposition. Therefore only the largest SEP events with orders of magnitude more energy flux would provide enough deposition to dominate the effects of solar photons. In a recent numerical study, Sheel et al. [2012] calculated the total ionization rates and electron density profile from the energy deposition profile obtained for the 29 September 1989 event. They find that this event can generate 10 4 cm 3 excess electrons at 30 170 km altitude, which is much larger than the typical values at altitudes below 100 km and thus should be detectable in the radio science experiments. They also find that an event two orders of magnitude less intense than the 20 September 1989 event are the smallest events able to produce sufficient numbers of electrons (3000 cm 3 )at80km altitude to be easily detected by radio occultations. Such events are also capable of causing attenuation sufficient to eliminate MARSIS surface reflections. In another modeling investigation focusing on the Mars Odyssey MARIE observations, Luhmann et al. [2007] investigated periodic modulations observed in the SEP measurements in the Odyssey spacecraft orbit. They suggested that the combination of magnetic field-aligned solar energetic particle anisotropies, instrument field-of-view restrictions, and SEP absorption by the obstacle presented by Mars and its atmosphere can account for the modulations observed in the data. Thus SEP observations in near Mars space, including within the ionosphere, may be altered by the presence of the planetary absorber and varying interplanetary magnetic field orientation. [7] All of these previous studies suggest that SEPs can affect the Martian ionosphere; however, the importance of SEP effects is still uncertain. Understanding these effects is important for determining the radiation conditions near Mars for future missions as well as for fully characterizing the impact of solar activity on the atmospheric evolution at Mars. In this study, we investigate the direct influence of SEPs on the Martian ionospheric electron density by combining the MAG/ER evidence of SEP event occurrence with the ionospheric observations from the MGS radio science experiment. In Section 2, we describe the data used in this work. Section 3 discusses a number of individual cases, and Section 4 presents the results from our statistical study of the ionosphere during and in the absence of SEP events. In the last section, we discuss the implications of the observations and summarize the conclusions from our study. 2. Data [8] The MGS Electron Reflectometer (ER) and Radio Science (RS) experiments are described in the detailed 2of19

investigation papers by the instrument providers, referenced below. The ER measures differential electron flux (#/cm 2 str s ev) at a rate varying between 2 and 48 s (depending on energy channel and telemetry rate) at 30 energy channels ranging from 10 ev to 20 kev. It has a field of view of 360 14 divided into 16 sectors [Acuña et al., 1992; Acuña et al., 2001; Mitchell et al., 2001]. Although the ER instrument is not designed to make space weather-related measurements, it can still detect solar storm-related particle events. Under normal conditions the ER instrument detects ionospheric and solar wind electrons, which typically exhibit a steep energy spectrum [Brain et al., 2012, Figure 2a]. Because of the steep falloff of electron flux with energy, the highest three energy channels (10 20 kev) of ER can often be used to estimate the instrument background levels. The count rate in these channels includes contributions from Galactic Cosmic Rays (GCRs; energy distribution peaks at 1 GeV), solar energetic particles (between 10 MeV and 1 GeV), and the 10 20 kev electrons the instrument was designed to measure. In periods for which there is no known solar activity near Mars, the background count rate is typically less than 15 c/s (varying between 6 and 12 c/s), caused primarily by cosmic rays. These high-energy particles penetrate through the instrument housing and most of the material surrounding the instrument generating spurious counts with nearly isotropic distribution and varying only modestly with the solar cycle. On some occasions, though, the count rate in these channels increases up to 10 2 10 4 c/s. Many of these times are well-correlated with known space weather events. During these events, SEPs can be distinguished from high-energy electrons by analyzing the shape of the energy spectrum from 10 to 20 kev in units of count rate [Brain et al., 2012]. When SEPs dominate the measurements, they produce the same signal (count rate) at all energy channels as they penetrate the instrument housing (>30 MeV) and are not affected by the electrostatic analyzer. 10 20 kev electrons, however, have a very steep energy spectrum. Therefore a flat spectrum from 10 to 20 kev channels indicates penetrating particles (>30 MeV) while any detectable slope shows a contribution from 10 to 20 kev electrons. Using this method, a total of 85 discrete events recorded by MGS/ER can be reasonably attributed to the presence of particle fluxes that penetrate the detector housing (>30 MeV) between May of 1999 and November of 2006 (for details of the ER response to highenergy particles and detection of the penetrating particles as well as the data set of varying solar events recorded by the ER at Mars, see Brain et al. [2012] and Delory et al. [2012]). Of these 85 SEP events, 41 were found to be temporally correlated with the CMEs detected by Large Angle and Spectrometric Coronagraph Experiment (LASCO) on the Solar and Heliospheric Observatory (SOHO) (for the list of these events, see Delory et al. [2012, Table 1]). [9] The radio science experiment (http://atmos.nmsu.edu/ PDS/data/mors_1102/ [Hinson et al., 1999; Tyler et al., 2001]) provides electron density altitude profiles between 80 and 200 km. Over the MGS mission (Sep 1997 Nov 2006), 5600 profiles were recorded. Because of the geometry of the experiment, all profiles are terminator occultations having a SZA range between 70 and 90. Moreover 96% of the profiles are confined to 60 86 N latitude band. There are fewer measurements in the southern hemisphere (4% of the profiles are in 65 70 S latitude band). These profiles generally show a main peak (F1 layer) at around 135 km, and a lower peak (E layer) in the range 105 120 km [Fox et al., 2008], which are also consistent with most of the occultation experiment observations from the Mariner 4, 6, and 7 flybys [Fjeldbo and Eshleman, 1968], the Mariner 9 orbiter [Kliore et al., 1972], and the Viking orbiters [Zhang et al., 1990]. The F1 peak is created from the solar photoionization by the solar extreme ultraviolet (EUV) radiation, and the E region is formed by solar soft X-rays and associated electron impact ionization. Most of the plasma around this secondary layer is produced by impact ionization from the highly energetic photoelectrons [Fox et al., 2008; Withers 2009; Mendillo et al., 2011; Lollo et al., 2012]. In the analysis of Mariner and Viking radio occultation observations, Zhang et al. [1990] and Hantsch and Bauer [1990] found that the global dust storm that occurred in the beginning of the Mariner 9 main mission heated the atmosphere and appears to have elevated the ionosphere by 20 30 km without significantly altering the shape of the ionospheric density profile. During the MGS mission, one global dust storm lasting for several months took place in the second half of 2001. The profiles on which we focus in this study are taken far from this storm period, and should therefore be free of global dust storm effects. Moreover, in our analysis we include the profiles only from the northern hemisphere and away from the strong crustal fields to exclude the effect of these sources. [10] In this analysis we also use the Solar Heliospheric Observatory (SOHO) [Judge et al., 1998] and the Geostationary Operational Environmental Satellites-10 (GOES-10) spacecraft data near Earth (GOES I-M DataBook by Space Systems-Loral, available at http://goes.gsfc.nasa.gov/text/ goes.databook.html). The Solar Extreme Ultraviolet Monitor (SEM) experiment on the SOHO spacecraft (the Solar and Heliospheric Observatory: a joint mission of European Space Agency, United States National Aeronautics and Space Administration) provides integrated flux at two wavelength bands; 26 34 nm (EUV flux) and 0.1 50 nm (XUV flux). In our study we used the 5 min averages of the SOHO/SEM fluxes from http://www.usc.edu/dept/space_science/semdatafolder/long/, courtesy of Dr. Didkovsky, Dr. Judge, and Dr. Ogawa. The Energetic Particle Sensor (EPS) on GOES-10 measures proton flux near Earth and in our analysis we used 5 min averages of proton flux at energies greater than 5 MeV, which is provided by the National Geophysical Data Center at http://goes.ngdc.noaa.gov/data/avg/. 3. Case Studies [11] Of the 85 SEP events identified using ER data, we identify 6 periods for which radio science profiles were also available. These 6 periods are also listed among the SEP events which are linked with specific CMEs detected by the Large Angle and Spectrometric Coronagraph on board the Solar and Heliospheric Observatory (LASCO) instrument on SOHO by Falkenberg et al. [2011, Table 1] and Delory et al. [2012, Table 1]. Table 1 summarizes these time periods and associated information about the SEPs and electron profiles recorded during these times. Along with these 6 SEP periods, a quiet-time event and the well known extreme 2003 Halloween event are also listed in Table 1 for reference (the first and last row, respectively). Figure 1 shows the relative locations of Earth, Mars and the Sun at the time periods listed 3of19

Table 1. Properties of the SEP Events Detected by the MGS/ER for Which There Are Also Radio Science Electron Density Profiles Event Date Data Used (d) SEPs Earth-Sun-Mars Angle (deg) Solar Condition Peak Count Rate at ER Log (c/s) SZA (deg) Average Parameters of RS Density Profiles Obtained During Each SEP Event Latitude (deg) Local Time First Peak (km) Second Peak (km) Quiet time 2001 03 06 10 30 Max x 72.3 84 4.6 133 112 1 2000 11 12 13 115 Max 3.2 85 64 2.8 140 116 2 2001 04 04 10 27 Max 3.9 71.5 84 8.2 135 113 3 2001 04 12 6 26 Max 3.9 72.2 84 8.5 135 113 4 2001 04 17 10 24 Max 2.6 72.5 84.5 8.7 135 113 5 2003 05 30 6 30 Max 2.6 78.2 69.5 14 139 113 6 2005 01 21 14 115 Min 3 74.3 78 7 137 114 Halloween 2003 11 13 x 30 Max 4.1 x x x x x in the table. For all these events, Mars is leading the Earth and it is very unlikely that they are on the same Parker spiral field line. [12] In the following paragraphs we investigate the correlation between the ionospheric electron density variations and the time profile of each SEP event. In order to determine the possible effects of SEPs, we attempt to distinguish the effects of SEPs from the known effects of EUV and XUV variations. To isolate XUV effects, we exclude the RS profiles recorded near the times of flares from our analysis. We then assume that the Chapman approximation and photochemical equilibrium are valid in the Martian ionosphere (i.e., the main and secondary peaks are produced mainly by solar EUV radiation (20 80 nm) and soft X-ray photons (<10 nm), respectively, and the electron density variation with respect to SZA is cos(sza) 1/2 [e.g., Hantsch and Bauer, 1990;Zhang et al., 1990; Fox and Yeager, 2009; Schunk and Nagy, 2000; Fox, 2004; Mendillo et al., 2011] and, as will be explained in detail in the following sections, at least partially account for the EUV variations that have non-flare sources as well (e.g., the 27 day variation due to solar rotation). The remaining variations are tested for correlation with SEP fluxes. 3.1. A Quiet Time Event (10 April 2000) [13] Before exploring the Martian ionospheric response to SEP events, in this section, we first present a case which is free of any solar activity influence. Through this case we intend to introduce the data and the method we will use in our SEP analysis, and obtain a representative case for the undisturbed conditions that we can refer to while identifying possible disturbances caused by SEPs in the following case studies. In other words, this case is representative for the features and variations of the RS profiles that occur at quiet times. [14] The six cases that we will analyze in the following sections are gradual SEP events lasting a few days each. Therefore for this quiet time case study, introduced in Figure 2, we analyze about 10 days of data. Each panel shows different spacecraft data recorded during early April 2001 in time series format (see the first row in Table 1). Figures 2a and 2b show the particle count rates, summed over all pitch angles, from GOES-10 and MGS/ER, respectively. For this case, the count rates in the highest energy channels of GOES-10 and ER are almost constant, indicating no solar activity during the time of interest. In the following case studies Figures 2a and 2b will help us keep track the time evolution of solar events at Earth and Mars. Figure 2c shows the ratio of the ER count rate from the highest two energy channels (12 kev/18 kev), which will help us distinguish the electron or proton dominated periods during the event (as explained in Data section). For this event, the ER count rate is 10 cc/s and the ratio of the highest two energy channels is close to 1, implying detection of GCRs but no high-energy electrons or SEP protons near Mars. In Figure 2d, the black line shows the proxy for the upstream magnetic field strength in nt estimated for each orbit using the MGS/MAG measurements (at 400 km). In this calculation, only the magnetic field data taken on the dayside (SZA < 110 ) in the magnetosheath (where ER electron data show sheath signatures [Mitchell et al., 2001]) are considered. After excluding the observations over crustal sources, remaining measurements are fit to a cos(sza) function and extrapolated to the field magnitude at SZA = 0 [Brain et al., 2005]. As this proxy is correlated with the upstream solar dynamic pressure, it is used to identify the shock arrival or ICME-related discontinuities in this study. (The proxy can be found at http://lasp.colorado.edu/mop/ people/brain/resources_files/proxies/subsolfield.html). On the same panel, green stars indicate the crustal field strength Figure 1. Relative positions of Earth and Mars with respect to the Sun (yellow dot) during the 6 SEP events recorded by MGS/ER and listed in Table 1. Gray, blue, pink, sky blue, black, green, brown, and red dots correspond to the Quiet time event, 6 solar events and the Halloween event, respectively. A typical Parker spiral interplanetary magnetic field is shown for reference (dashed green lines). 4of19

Figure 2. Time series data detected during the Quiet time event listed in Table 1. (a) GOES particle count rate recorded at Earth. (b) MGS ER background count rate (black curve). (c) The ratio of the MGS ER count rate from the highest two energy channels (d) Upstream magnetic field proxy from MGS (dark green curve) and the crustal field strength (green stars) obtained at the location of RS profiles at about 400 km altitude from the spherical harmonic expansion model of crustal sources by Cain et al. [2003]. (e) EUV flux proxy at Mars. (f) X-ray data from SOHO recorded at Earth. The dotted vertical lines in all panels indicate the times of RS profiles. The colored solid vertical lines in Figure 2b show the RS profiles that are grouped in a few days apart. at the geographical location of the RS profiles (Table 1) at about 400 km altitude obtained from the spherical harmonic expansion model of crustal sources by Cain et al. [2003]. As seen here, crustal fields are small over the region where the RS profiles are taken and therefore should have almost no effect on the observations. [15] Figure 2e shows the F10.7 radio flux proxy for Mars, an estimate for the EUV flux at Mars (flux in 10 22 Wm 2 Hz 1 at wavelengths 2 100 nm). This proxy is obtained by distance scaling (1/r 2 ; r is the radial distance from the Sun) and timeshifting (A 26-day solar rotation rate is assumed) of the F10.7 radio flux measured at Earth [Mitchell et al., 2001]. (The proxy can be found at http://sprg.ssl.berkeley.edu/brain/proxies/ euvproxy.html) Figure 2f shows the 26 34 nm EUV and 0.1 50 nm XUV flux detected by SOHO at Earth. As the angle between Mars and Earth is small (30 as indicated by gray dots in Figure 1; i.e., both planets see the same side of the Sun), the solar EUV intensity time profile at Mars is expected to be similar to what is observed at Earth, only with a 5min delay. During the time period in Figure 2, the XUV time profile is almost constant and there are no detectable solar flares. As reported by Mendillo et al. [2006] and Mahajan et al. [2009], the ionospheric density increases in response to flare photons and this density elevation effect lasts about an hour. Since the present work seeks to isolate the effects of particles, and not photons, we exclude from our analysis all RS profiles that occur within the first two hours after an XUV flare. For this case there are no flare-affected profiles; however in the following case studies we do have such profiles and the data in Figure 2e help us to detect and exclude the flare-affected RS profiles from our analysis. The vertical dotted black lines indicate the times of all 65 RS profiles recorded during the time interval in Figure 2. The profiles occur as frequently as every two hours (MGS orbited Mars every 118 min). [16] In order to investigate the response of the Martian ionosphere to the varying plasma environment observed during a SEP event near Mars, we will group the RS profiles considering their times with respect to the SEP event. Then we will compare the average profiles from each group, focusing on different altitudes. For this case there is no solar event that we can refer to in grouping the profiles but still we group the profiles a few days apart to reveal the variations of ionospheric electron density under undisturbed conditions. In grouping the profiles we use the same color convention throughout the text. The profiles recorded before and after the event under quiet time conditions (when ER is dominated by GCR) are marked in tones of gray. The profiles detected when ER measurements are dominated by SEPs are in red. The profiles that are detected when ER is dominated by high-energy electrons (10 20 kev) are in cyan if they are before the event peak (when the pressure pulse from any passing CME has not yet arrived), blue if they are after the event peak (when SEP and pressure effects could be mixed). In Figure 2b, the solid vertical lines show the grouping of the profiles for this quiet time case. The profiles are in different tones of gray as they are 5of19

Figure 3. Grouping of the 65 RS electron density profiles recorded during the quiet time event in Table 1 (refer to Figure 2b for grouping): light gray, days 1 3; lightest gray, days 4 6; dark gray, days 7 9. (a) Comparison of averaged electron density profile of each group. (b) Zoomed in view of the region near the main peak. (c) Zoomed in view of the region near the second peak. (d) Zoomed in view of the region below 100 km altitude. recorded when ER is dominated by GCR. Figure 3a shows the averages with standard deviations, and Figures 3b 3d compare these averages focusing on different altitudes. [17] The electron density profiles of the Martian ionosphere have been identified as Chapman-like [e.g., Zhang et al., 1990; Fox and Yeager, 2009]. The main peak is produced by photoionization due to solar EUV radiation (20 80 nm) and is located between 120 and 140 km altitudes depending on the SZA. The variation of the electron density with respect to SZA and EUV flux has been found to be cos(sza) 1/2 and (EUV flux) 1/2 [e.g., Hantsch and Bauer, 1990; Zhang et al., 1990; Fox and Yeager, 2009]. A secondary peak due to mostly soft X-ray photons (<10 nm) is located between 90 and 120 km altitudes and most of the plasma at this altitude is produced by impact ionization by the energetic photoelectrons created during photoionization [Schunk and Nagy, 2000; Fox, 2004; Mendillo et al., 2011]. The mean value of this peak and its electron density dependence on SZA (cos (SZA) 1/2 ) is well predicted by the Chapman theory [Mendillo et al., 2011]. Even though Chapman representation provides a useful insight about the real state of the ionosphere as a first order approximation, significant variation of the secondary peak suggests that the photochemical processes that dominate at this altitude are more complex than the Chapman theory. Substantial variability of solar soft-x-ray irradiance on short time scales and the lack of accurate knowledge of the irradiance at these wavelengths make the identification of the features of this layer even harder. As seen in Figure 3a, the profiles recorded in Figure 2 have the main peak (1.0E5 9% cm 3 ) at around 133 km and the secondary peak (4.9E4 7% cm 3 ) at around 112 km altitude. [18] The range of SZA sampled during this time period (72.2 72.5 ) is very small; therefore we assume the RS profiles are free of SZA variation effects. In general, for the RS profiles that are recorded in about 10 15 consecutive days, SZA does not change much and therefore has a small effect on the density variation (assuming the dependence of density on SZA is cos(sza) 1/2 ). Moreover, during quiet times, EUV flux does not vary substantially in about 10 days and in general the variation trend in the main peak density is consistent with the EUV flux (by (EUV flux) 1/2 ). For this case EUV flux variation ((EUV flux) 1/2 ) is about 7% at most (changes from 51.0E 22 Wm 2 Hz 1 to 58.9E 22 Wm 2 Hz 1 in Figure 2e), and in Figure 3b the variation in the main peak averages is lower than 4% (from 1.01 6% cm 3 to 1.05E5 8% cm 3 ). Note also that XUV in Figure 2e is almost constant during the time period of interest. Even though XUV flux (0.1 50 nm) is a loose proxy to estimate the soft xray flux (<20 nm), it still implies that the soft xray time profile is likely stable in the time period we consider. In Figure 3c, the variation in the second peak averages is about 10% (4.5 7% cm 3 -light gray and 5.0 7% cm 3 -dark gray), which is within the expected range (about 5 10%) due to longitude changes [Haider et al., 2006]. Moreover, we do not observe a significant variation in the electron density averages below 100 km altitude. The variations observed in the profiles plotted in Figure 3 are representative of many quiet time cases analyzed using the same method. In other words, the variation in the main and second peak within a period of about 10 days, when there is no observable solar activity, is not more than 10%. In the following cases we will explore the SEP effects using similar data and method during the 6 solar events listed in Table 1. 3.2. Event of 11 November 2000 [19] Similar to Figure 2, Figure 4 shows the time series data for the first SEP event in Table 1. This event lasts about 13 days, during which the ER background count rate increases by 3 orders of magnitude from the quiet time count rate due to GCRs. For this case Mars is 115 ahead of Earth which reduces our ability to predict the solar radiation effects accurately. Moreover, the RS profiles for this event are taken at high SZA (the highest SZA of all 6 events in Table 1) and therefore have the lowest ionospheric densities with highest standard deviations (bigger error bars compared to other cases). Small variations and large standard deviations in the electron density also prevent reliable identification of density variations which stay within the error bars. We nevertheless include a brief analysis of the event below as we find it interesting and helpful in understanding the SEP effects at Mars. [20] For this case, Mars is 115 ahead and 0.5 AU downstream of Earth. Therefore the difference in appearance of the SEP event time profiles detected at Earth by GOES and at Mars by MGS in Figures 4a and 4b may relate to different magnetic connectivity to the shock source of SEPs at the two locations. Moreover, CME shock arrival is not clear in the magnetic field data (Figure 4d), which suggests that the geometry of the event is likely as in CASE B in Figure 5. (The 6 SEP events that we study in this paper are in the catalog of energetic particle events at Mars associated with solar flares and coronal mass ejections in Delory et al. [2012]. For more detailed information about the identification of the events and their onset times at both planets please refer to Delory et al. [2012] and 6of19

Figure 4. As in Figure 2, but for the first event in Table 1. The colored solid vertical lines in Figure 4b show the RS profiles with respect to the SEP event: light gray, before the event; red, at the time of SEP protons; blue, at the time of high-energy electrons (10 20 kev) after the peak of the event; and dark gray, after the event. The black vertical lines in Figures 4b and 4f show the RS profiles that occur within the first two hours after an XUV flare. Falkenberg et al. [2011]). Until 11 09/04:00, ER count rate is about 10 cc/s, indicating dominance of GCRs in the measurements. An increase in the ER background (also a flat count rate ratio in Figure 4c) implies that SEPs dominate between 11 and 10/00:00 and 11 13/00:00 UT. Often associated with the passage of the CME shock, softening of the energy spectrum of the plasma and an increase of the 10 20 kev electron flux is evident in Figure 4c. After 11 13/ 00:00 the count rate ratio starts to increase and 10 20 kev electrons dominate the ER measurements. At around 11 18/ 20:00, ER count rate drops to its quiet time values (dominated by GCRs again) indicating the end of the event. [21] Figure 4e shows the F10.7 radio flux proxy for Mars, and Figure 4f shows the 26 34 nm EUV and 0.1 50 nm XUV flux detected by SOHO at Earth. The large angle between Mars and Earth during this event makes it hard to use this data reliably. Indicated by the vertical lines in all panels, for this event, we analyze 129 RS electron density profiles which are far from strong crustal fields and free of crustal field effects (Figure 4d). As explained in the previous section in order to investigate the response of the Martian ionosphere to the varying plasma environment observed during a SEP event, we group the RS profiles considering their times with respect to the SEP event and mark them in different colors as shown in Figure 4b. The gray profiles are observed during quiet times before and after the peak of the event. The red profiles are detected at the times of SEPs and the blue profiles are recorded when ER measurements are dominated by highenergy electrons after the event peak. Even though the angle between Earth and Mars is large (115 deg.), we still assume that the large flare occurring at 11 10/02:00 UT may have an effect at Mars and exclude the RS profiles during these times (marked in black in Figures 4b and 4f). Figure 6a compares the average profiles from each group along with the corresponding standard deviation of the group, while Figures 6b 6d focus on these average density profiles at different altitudes. Figure 5. Possible geometry of the events listed in Table 1. Relative positions of the Sun, Mars and the CME shock. (Figure adapted from Reames [1999].) 7of19

Figure 6. As in Figure 3, but for the first event in Table 1: light gray, before the event; red, at the time of SEP protons; blue, at the time of high-energy electrons (10 20 kev) after the peak of the event; and dark gray, after the event. [22] As seen in these figures, the main peak is at 140 km (5.4E4 15% cm 3 ) and the second peak is at about 116 km (2.4E4 24% cm 3 ). Unfortunately, for this case, due to the large angle between Mars, Earth and the Sun and the substantial uncertainties in the electron density (one reason may be high SZA of the profiles, 85 ), we cannot derive any variability information accurately at both peaks. Even considering the fact that the observation statistics are not good enough to distinguish any variability, it is still interesting that the density is anomalously low near the second peak at around 110 km (2.14E4 20% cm 3 for the red curve). As we will show in the following two strong cases (second and third events in Table 1) in more detail, this decrease coincides with the increasing SEPs and cannot simply be explained by the usual variation of density at this altitude. Due to the lack of data at low altitudes below 95 km for three of the four profile groups, we cannot explore the electron density variation at the very bottom of the ionosphere for this case. For the one profile set for which we have data below 95 km (0.64E4 33% cm 3, dark gray at quiet times after the event), the electron density is slightly higher than the typical values we observe at the quiet times (Figure 3d). One reason for such relatively high electron density at low altitudes may be the complicated behavior of the time-dependent interaction at disturbed times and/or precipitation of particles related to the solar wind interaction itself near the terminator. 3.3. Event of 4 April 2001 [23] Similar to previous cases, Figure 7 shows different spacecraft data as a time series recorded during early April, 2001, the second SEP event in Table 1. This event is the strongest of those listed in Table 1 after the Halloween event, and the angle between Earth and Mars was small (<30 )at this time, therefore we expect that any influence of SEPs on Figure 7. As in Figure 2, but for the second event in Table 1. The colored solid vertical lines in the Figure 7b show the RS profiles with respect to the SEP event: light gray, before the event; cyan, at the time of high-energy electrons (10 20 kev) before the peak of the event; red, at the time of SEP protons; blue, at the time of high-energy electrons (10 20 kev) after the peak of the event; and dark gray, after the event. The black vertical lines in Figures 7b and 7f show the RS profiles that occur within the first two hours after an XUV flare. 8of19

Figure 8. As in Figure 3, but for the second event in Table 1: light gray, before the event; cyan, at the time of high-energy electrons (10 20 kev) before the peak of the event; red, at the time of SEP protons; blue, at the time of high-energy electrons (10 20 kev) after the peak of the event; and dark gray, after the event. the Martian ionosphere should be most pronounced and most easily compared to simultaneous terrestrial observations. During this event, the count rates in the highest energy channels of GOES-10 and ER increase about 3 orders of magnitude from their nominal value (Figure 7b). The event lasts for 5 days, consistent with a gradual SEP event. Crustal fields are small over the region where the RS profiles are taken and therefore the effects are expected to be negligible and ignored in the analysis. [24] The difference in appearance of the SEP event time profiles detected at Earth and Mars in Figures 7a and 7b may be due to different magnetic connectivity of both planets to the shock front. Moreover, the different instruments at each planet give different views of the plasma event, even if the photon events are quite similar. In Figure 7b, the ER count rate before 03 31/18:00 UT is 10 cc/s implying detection of GCRs for which the ratio of the highest two energy channels is about 1.3. At 04 01/18:00 UT both ER count rate and the count rate ratio start to increase, which indicates dominance of 10 20 kev electrons at the beginning of the SEP event. At around 04 03/02:00 UT, even though the count rate continues to increase the count rate ratio starts to decrease, suggesting that the SEPs (>30 MeV) dominate the high-energy electrons (10 20 kev). And after 04 04/ 22:00 UT CME-related 10 20 kev electrons start dominating the measurements again. The inferred 10 20 kev flux increase (energetic storm particles i.e., ESPs) near the center of the MGS ER event is coincident with an increase in the magnetic field which may represent the interplanetary CME initial shock arrival. The schematic of the possible geometry of the event and the relative position of Mars is shown in Figure 5 (CASE A). As mentioned before, the passage of the CME shock is often associated with softening of the energy spectrum of the plasma and an increase in the 10 20 kev electron flux. This can be seen in the ER data as a nonzero slope in the 10 20 kev energy spectrum. After 04 07/ 04:00 UT, the count rate ratio falls back to 1.2 implying returning back to normal plasma conditions and measurements of GCRs at these high-energy channels. [25] For this particular case the angle between Mars and Earth is small and therefore solar radiation profiles at Mars should be similar to what is observed at Earth (only with a 5 min delay) allowing us to use the data in Figures 7e and 7f reliably. During the time period shown in Figure 7, the XUV time profile has a negative slope in general; however, there are a few instances where we observe solar flares lasting minutes to hours. We exclude from our analysis all RS profiles that occur within the first two hours (marked in black in Figures 7b and 7f) after an XUV flare. [26] The RS profiles that are grouped considering their times with respect to the SEP event are shown in Figure 7b. The gray vertical lines indicate the profiles obtained before and after the event. Cyan indicates the profiles recorded when 10 20 kev electrons dominate the ER measurements in the very beginning of the event, the red lines indicate the profiles recorded when SEPs dominate the measurements (presumably protons, as the gradual proton events are dominated mostly by ions, protons >30 MeV) and the blue lines show the profiles when CME related high-energy electrons become dominant again after the first arrival of SEPs and the CME shock. Figures 8a 8d compare the average profiles from each group, focusing at different altitudes. The main peak is at around 135 km (1.2E5 6%) and the second peak is at 113 km (5.3E4 9% cm 3 ). In Figure 8b, even though the density variations are within the uncertainty ranges, it is still consistent with the EUV flux proxy plotted in Figure 7e. The EUV averages are 96.8, 92.6, 89.1, 78.3, 72.8, and peak density averages are 1.19E5 5% cm 3, 1.18E5 5% cm 3, 1.16E5 8% cm 3, 1.13E5 6% cm 3, and 1.11E5 5% cm 3 for the light gray, cyan, red, blue, and dark gray profiles, respectively). [27] The variation at and below the secondary peak, however, is hard to explain. The density at 113 km during the rise of the SEP event (red curve) is anomalously low (variation in the density from cyan (6.1E4 9% cm 3 ) to red (4.5E4 5% cm 3 )is26%). Even though the density variation between the interior curves (gray and blue) is within the uncertainty limits, the red (cyan) curve is out of the uncertainty range of the cyan (red) curve confirming that the anomalous depletion is real. Note that we have already eliminated the flare effects. The XUV variation is very small, implying that the solar irradiance at <20 nm is likely to be constant during the event. Moreover, during the event SZA ranges from 71.8 to 72.1 and therefore has also a negligible effect on the profiles. In the previous quite-time event, the variation of the RS profiles below 100 km occurred under similar solar irradiance and SZA conditions. We have found that the variation in the second peak is typically 10%, which is lower than the observed variation in this SEP event. Moreover, Haider et al. [2006] show that the electron density variation due to longitude differences at both peaks can be at most 5 10%, which is again smaller than the variation we observe here. Therefore the observed 25% density decrease cannot be explained by the usual variability of the second peak. [28] Assuming that the Martian atmosphere is composed of CO 2 only (95%), we calculate the penetration depth of 9of19

Figure 9. Penetration depth of electrons and protons in CO 2. (The plot is obtained using the tables at http://www.nist.gov/ pml/data/star/index.cfm, assuming that the number density of CO 2 is N/cm 3 =5.88E18exp( z/7.00) + 3.55E13 exp ( z/16.67), where z is altitude in km, from MTGCM by Steve Bougher, which was run for Ls = 270, F10.7 = 105 at Mars and published at http://www.issibern.ch/teams/martianplasma.) 30 MeV protons as shown in Figure 9: In this calculation, we obtain the number density of CO 2 using N/cm 3 =5.88E18 exp ( z/7.00) + 3.55E13 exp ( z/16.67), (z is altitude in km) from Mars Thermosphere General Circulation Model by Steve Bougher, which was run for Ls = 270, F10.7 = 105 at Mars and published at http://www.issibern.ch/teams/martianplasma. Then we determine the penetration depth of electrons and protons in CO 2 as a function of energy using the tables provided by the Physical Measurement Laboratory of the National Institute of Standards and Technology at http://www.nist.gov/ pml/data/star/index.cfm. As seen in Figure 9, we find the penetration depth of 30 MeV protons to be 30 km. Therefore it is unlikely that >30 MeV SEPs affect the ionosphere directly as they deposit their energy well below ionospheric altitudes. However, 30 MeV SEPs measured by MGS are certainly accompanied by lower energy SEPs that may deposit their energy at altitudes measured by the radio science experiment. In fact, Leblanc et al. [2002] and Sheel et al. [2012] explored the influence of a typical SEP event on the Martian ionosphere by test particle simulations where they find that protons having energies greater than a few tens of kev are not deflected by any magnetic field near Mars and do not charge exchange. These high-energy particles deposit their energy into the Martian atmosphere before they reach these low altitudes (1 10 kev protons deposit their energy between 80 and 120 km altitudes), likely causing extra heating or ionization in the ionosphere. Electron density profiles estimated by Sheel et al. [2012] show electron densities in excess of 3000 10 4 cm 3 below 100 km altitude due to SEPs, while Leblanc et al. [2002] predicted that unless the SEP event is strong enough, it will not produce sufficient ionization to rival the EUV effects. One reason for this inconsistency may be because the energy deposition due to solar EUV/UV estimated by Sheel et al. [2012] below 120 km altitude is much smaller than the observations reported by Fox and Dalgarno [1979] and Leblanc et al. [2002]. Our findings here agree with the results by Leblanc et al. [2002] in the sense that the extra ionization created by typical SEPs between 90 and 120 km is not high enough to be detected by RS experiments. In fact, surprisingly, in contrast to the expectation of an ionospheric density increase, we observe a decrease (25%) between 90 and 120 km. The observed depletion may be due to a process associated with the SEPs, the mechanism of which is not known to us. Withers [2009] reported four examples of reduced electron density profiles which are also mysterious. The dates and occurrences of these profiles are different from the ones we analyze in this study [Withers, 2009, Figure 25]. They report an unusually reduced density for one out of many profiles occurred under similar conditions (1 out of 25 profiles). The reduced densities imply low production or high loss rates just for this one profile; however, the main cause of these anomalous profiles is not known yet either. [29] The absence of an increase in the electron density observations between 90 and 120 km (for the SEPs profiles, i.e., magenta curve) may be because most of the deposited energy goes to heating rather than ionization and/or the ionization is at altitudes below 90 km. The Leblanc et al. [2002] calculations were limited to above 100 km altitude; however, Sheel et al. [2012] extend their calculations down to the deep ionosphere and also predict a third peak at around 65 km altitude which is larger than the second peak at 115 km altitude. Unfortunately we cannot investigate this third peak during SEPs in our analysis as the lowest altitude of the RS profiles is 90 km for this case and 80 km in general. [30] Our finding of density depletion is also in contrast to the indirect measurements of electron density in the Martian ionosphere by the MARSIS radar sounder. Lillis et al. [2010] reported about 20 200% increase in the subsolar TEC recorded by MARSIS at the time of two solar events (July 2005 and December 2006; i.e., different from the events we analyze here). Moreover, Gurnett et al. [2005], Morgan et al. [2006, 2010], and Espley et al. [2004] reported ionospheric absorption of the radar signals reflected from the surface, which last several days and are correlated with the SEP events detected at Mars by MGS ER. These MARSIS blackouts suggest increased ionization in the Martian ionosphere at the times of SEPs. Considering these observations, we expect that RS profiles should reflect an electron density increase induced by SEPs; however, our careful search of the profiles does not reveal such a response in the Martian ionosphere between 90 and 120 km for our events. As discussed before, one reason for both the TEC increase and the MARSIS blackouts may be because the extra ionization produced by SEPs is below 90 km. Another reason for the detected increase in the TEC by Lillis et al. [2010] may be solar flares whose effects are not excluded in their analysis. In Lillis et al. [2010], there are other instances that exhibit elevated TEC but no SEP signatures observed at Mars by MGS ER [Lillis et al., 2010, Figure 3]. As also recommended by Lillis et al. [2010], more detailed temporal analysis with respect to solar flares and SEPs (as we did here) is needed to reveal the direct SEP effects at Mars. Knowledge of the entire SEP energy spectrum might be helpful in predicting their effects but unfortunately with the limited observations from the ER instrument we can determine neither the flux of the penetrating particles nor the entire energy spectrum [Delory et al., 2012]. [31] The density variation at the lowest altitudes (<100 km) during the time periods when 10 20 kev electrons dominate the ER measurements is also interesting to note (Figure 8d). The highest electron density (1.3E4 20% cm 3 at 90 km) at these low altitudes is observed for the cyan curve implying that this is an effect of the 10 20 kev electrons in the beginning 10 of 19