NEAR-INFRARED BRIGHTNESS OF THE GALILEAN SATELLITES ECLIPSED IN JOVIAN SHADOW: A NEW TECHNIQUE TO INVESTIGATE JOVIAN UPPER ATMOSPHERE

Our observations were conducted with the Infrared Camera and Spectrograph (IRCS) (Kobayashi et al. 2000) using adaptive optics (AO188; Hayano et al. 2010) with the natural guide star mode on the Subaru telescope (Iye et al. 2004), the Wide Field Camera 3 (WFC3; MacKenty 2012) on the Hubble Space Telescope, and the Infrared Array Camera (IRAC; Fazio et al. 2004) on the Spitzer Space Telescope (Werner et al. 2004). Our observations are summarized in Table 1. Since our observational targets are specific astronomical events, the opportunities for observation are limited and timing is absolutely critical. The separation from the Jovian limb to the satellites is less than 1 arcmin in our observations, thus the greatest difficulty results from stray light due to the close proximity of Jupiter to the target satellites. Therefore, we developed an advanced method for our observations. Although details of our method depend on each observation, a basic strategy is as follows. Jupiter was kept out of the detector field of view (FoV) during the observation to avoid electric crosstalk on the detector array and stray light from Jupiter, which is ∼10⁸ brighter than the eclipsed satellites. Observations were conducted with non-sidereal tracking on Jupiter (outside of the FoV) to fix the stray light pattern on the detector during the observation.¹³ This minimizes the systematic error due to the Jovian stray light. Owing to absorption by methane in the Jovian atmosphere, Jupiter is dark in methane bands (CH₄-long filter for Subaru/IRCS, F139M for Hubble/WFC3, and Channel 1 for Spitzer/IRAC), so observations in the methane bands are very effective for reducing the stray light from Jupiter. Short integration times were required to avoid smearing of the target satellite owing to the relative movement of the target satellites to Jupiter. Allowing the eclipsed satellite move relative to the detector using Jupiter tracking has the advantage of effectively dithering the observations so that we can average out any detector issues. The satellite ephemerides are provided with great precision by the Natural Satellites Ephemeride Server MULTI-SAT¹⁴ (Emel'Yanov & Arlot 2008) and the JPL-HORIZONS¹⁵ (Giorgini et al. 1996). The size and location of the eclipsed satellite are accurately known and are therefore easily differentiated from detector effects.

In addition to the standard data reduction such as dark frame subtraction, flat-field correction, and calibration using standard stars, we conducted a special data processing described below.

In each observation we obtained a time series of the observed images, in which the target satellite in eclipse was moving in the fixed stray light pattern from Jupiter outside of FoV. The stray light pattern for each single frame was constructed from other images in which the target satellites moved more than the satellite size (0.8–1.6 arcsec) from the single frame, and was subtracted. In the case of observations with the methane bands, the stray light level from Jupiter is equal to or less than the sky level, so the uncertainty due to the stray light subtraction is also equal to or less than the uncertainty from the photon noise from the sky level. In the case of observations with non-methane bands (J-band filter for Subaru/IRCS and F160W for Hubble/WFC3), the stray light level from Jupiter was 5 to 12 times brighter than the sky level, so the uncertainty from the photon noise from the stray light was added to that from the sky level. However, the stray light pattern from Jupiter was successfully canceled out due to the observations with Jupiter tracking; thus, systematic uncertainty from the stray light subtraction should be negligible. Another Galilean satellite out of eclipse (∼10⁶ brighter than the eclipsed satellites) was in the images in some data (Subaru/IRCS data in 2012 February 21 and 2012 July 26, and Hubble/WFC3 data in 2013 February 5 and 2013 April 8), and the target satellite in eclipse was contaminated by the nearby bright satellite. Since the satellites of the contamination source had different movements relative to the target satellite in eclipse and Jupiter, a contamination pattern was not fixed in the images. Even in such a case, since the contamination pattern does not have a local structure, the contamination pattern was evaluated around the target satellite and subtracted locally.¹⁶

After the stray light subtraction, brightness of the target satellite in eclipse was evaluated. Since we know the exact position and diameter of the satellites, and we can correct the positioning information in the FITS header of each image by comparing it to field stars and the satellite out of eclipse in the images, we can find the pixels where the target satellite in eclipse should exist even in the case of non-detection. The amount of positioning correction by comparing to field stars was less than five pixels. The total uncertainty after the subtraction of the stray light from Jupiter and the satellite out of eclipse was evaluated from the deviation of the pixels around the target satellite in eclipse.

One possible explanation is (1) the atmospheric emission from the satellites in eclipse. Auroral emission from electron excited atomic oxygen at UV (130.4 and 135.6 nm) (Hall et al. 1998; McGrath et al. 2013) and optical (630.0 and 636.3 nm) (Brown & Bouchez 1999) at both poles of Ganymede in eclipse was known, but this auroral emission cannot explain the uniformity of brightness as shown in Figure 1. Thus, the origin of this brightness should be radiation from whole atmosphere of the satellites. In the case of the Earth atmosphere, OH molecules are the main carrier of the airglow emission in the near-infrared wavelengths, which are excited by sunlight at dayside and emit photons at nightside. The Galilean satellites are also expected to have OH molecules in their atmosphere, since OH molecules can be generated from water ice on their surfaces by sputtering due to interaction with the plasma particles around Jupiter. Such OH molecules in their atmosphere can be excited by the sunlight while out of eclipse and emit photons during their eclipses. The column density of OH is estimated to be 10¹⁰–10¹¹ cm⁻² in the Ganymede atmosphere (Marconi 2007), which is comparable to that of the OH layer in the Earths atmosphere. However, the Ganymede SED, especially the non-detection at 3.6 μm, is difficult to explain by OH airglow because there is an emission band (Δν = 1) at 2.5–3.5 μm (Stair et al. 1985) as shown in Figure 3, although the excitation state of OH molecules and chemical environment in the atmosphere of the Earth and the Galilean satellites are different, resulting in different spectral shapes. The density and composition of Europa's atmosphere are very similar to those of Ganymede and Callisto, therefore it is difficult to explain the large difference in observed brightness among the three. Additionally, atmospheric emission was detected from Europa by the Cassini spacecraft during a fly-by, but not from Ganymede (Porco et al. 2003), which is inconsistent with this explanation.

The second candidate is (2) illumination by emission from the dark side of Jupiter. Lightning (Dyudina et al. 2004), aurora (Kim et al. 1993; Gladstone et al. 2007; Radioti et al. 2011), and nightside airglow (Gladstone et al. 2007) of Jupiter are known and they can be candidates of the illuminator. However, it is difficult for these illuminators to explain the fact that the brightness of Europa is much darker than that of Ganymede and Callisto because Europa is located closer to Jupiter, thus Europa should be illuminated more strongly from Jupiter than the others. In addition, the brightness of the satellites in eclipse was kept during our observations (more than one hour in some observations), thus it is also difficult to be explained by sporadic events like lightning and aurora.

We estimated the effect of (3) illumination by the reflected sunlight from the other satellites. Assuming that Io out of eclipse as an illuminator is located 3 × 10⁵ km away from Europa in eclipse,¹⁸ the brightness of Io seen from Europa (with a phase angle of 0°) is ∼6× 10⁶ of that seen from the Earth. Since extinction owing to phase angle (Sun–Io–Europa) is >10² when the phase angle is >135°¹⁹ (Simonelli & Veverka 1984), the brightness of Io seen from Europa with the phase angle of 135°becomes ∼6 × 10⁴ of that seen from the Earth. In addition, the brightness of the Sun seen from Jupiter (or Io/Europa) is ∼4× 10⁻² of that seen from the Earth. We know that the brightness ratio between the Sun and Io seen from the Earth is ∼1× 10¹³, then we obtained their brightness ratio seen from Europa as <10⁷.

This effect cannot explain the detected brightness of Ganymede and Callisto, whose brightness was ∼10⁻⁶ of that out of eclipse, but it can explain the detected brightness of Europa, whose brightness was ∼10⁻⁷ of that out of eclipse. In fact, when we detected the Europa brightness in eclipse on 2014 March 26, the phase angle was 134° and the distance between Europa and Io was 3.4 × 10⁵ km. These values are almost same as the assumed values in the estimation, meaning that Europa in eclipse was illuminated from Io in the most efficient configuration on this day. On the other hand, Europa in eclipse was not detected with the same filter on 2013 April 8. On this day, the distance between Europa and Io was 2.8 × 10⁵ km and the phase angle was 155°. Io appears brighter by a factor of ∼1.5 by the distance effect, but darker by a factor of >6 by the phase angle effect (Simonelli & Veverka 1984) than that on 2014 March 26. In the same way, the non-detections of Europa on the other days can be explained by the configuration of the Jovian system. Thus, illumination by the reflected sunlight from the other satellites explains both the brightness of Europa on 2014 March 26 and the non-detection in the other days.

We propose a hypothesis that (4) the Galilean satellites are illuminated by sunlight forward-scattered by hazes in the Jovian upper atmosphere to explain the brightness of Ganymede and Callisto in eclipse. A detailed discussion for this model is found in T. Nakamoto et al. (2014, in preparation), which details that illumination is dominated by hazes around the top of the stratosphere of the Jovian atmosphere. A methane absorption-like feature at 1.4 μm is found in the Ganymede SED in eclipse, as in the Jovian spectrum (Rayner et al. 2009) as shown in Figure 3. A methane abundance around the top of the stratosphere is almost same as that at higher pressure regions in the Jovian atmosphere (∼10⁻³ in mole fraction) (Moses et al. 2005), which can explain the 1.4 μm methane-like absorption feature in the Ganymede SED in eclipse if this model is true. The albedo of Ganymede at >3 μm is much smaller than that around 1.5 μm due to water ice on its surface (Calvin et al. 1995), which also explains the non-detection of Ganymede in eclipse at 3.6 μm. In addition, the fact that Europa in eclipse was much darker than the others can be explained by this model because the effect of the scattered sunlight should be less at a satellite closer to Jupiter, and the depth of the observed Europa eclipses were deeper (impact parameters were smaller) than those of Ganymede and Callisto. In such a case, however, the extinction of Callisto (∼2 × 10⁻⁶) should be greater than Ganymede (∼4 × 10⁻⁶) because Callisto is located farther from Jupiter than Ganymede, but this was not the case. The difference of albedo and diameter are canceled out in this extinction comparison. This might be explained by the difference of the haze abundance in the Jovian upper atmosphere because the brightness of satellites in eclipse should depend on where (longitude and latitude) the sunlight was scattered in the Jovian atmosphere. The haze abundance in the Jovian atmosphere depends on position (Zhang et al. 2013), and Callisto might be illuminated by sunlight through the haze-rich region of the Jovian upper atmosphere when we observed it. As shown in Figure 2, all Ganymede and Callisto eclipses we observed were shallow (impact parameter >0.7), thus they may become darker at deeper eclipses in this hypothesis, which will be tested by our future observations.

As the transmission spectrum of the Earth atmosphere was measured from lunar eclipse observation (Pallé et al. 2009; Vidal-Madjar et al. 2010; García Muñoz et al. 2012), spectra of the Galilean satellites in eclipse could be tracers of the Jovian upper atmosphere, especially the abundance of haze and/or methane along the Jovian altitude, if this model is correct. In particular, secular change and location dependence of the Jovian atmosphere can be monitored by long-term observations of a number of eclipses. For example, observed brightness variances of Ganymede and Callisto shown in Figure 2 might be explained by the variance of the haze abundance in the Jovian atmosphere. Similar studies were conducted to investigate the composition of the Jovian atmosphere using Galilean satellites' eclipses (Smith et al. 1977; Greene et al. 1980; Smith 1980; Smith & Greene 1980). However, these previous works were based on ingress and egress eclipse light curves that probed reflected/refracted sunlight through regions of the Jovian upper atmosphere (specifically, the bottom of the Jovian stratosphere at pressures of several hundred mbars). This cannot explain the brightness we observe during total umbral eclipses.

To explain this brightness in the total umbral eclipse, T. Nakamoto et al. (2014, in preparation) propose forward-scattered sunlight by hazes in the upper region of the Jovian atmosphere (specifically, the top of the Jovian stratosphere at pressures of several to several tens of mbars). This pressure range has been poorly observed by previous spectral observations and in-situ observations by the Galileo probe. For example, in previous works, Zhang et al. (2013) explicitly stated that 10 mbar is the top of the sensitivity region, and Banfield et al. (1998) stated that the limit is 20 mbar. Our observation has the potential to resolve the vertical structure of hazes well above this pressure level if this model is correct. The Jovian atmosphere of an even higher region (<1 mbar) was investigated by observations of field star occultation by Jupiter (Raynaud et al. 2004; Christou et al. 2013), but occultations of field stars by Jupiter are much rarer than the Galilean satellite eclipses. Therefore, observations of the Galilean satellite eclipsed in the Jovian shadow can be a very unique method to investigate the Jovian upper atmosphere if this hypothesis is correct. The clouds and hazes are mainly produced in this pressure range (West 1988; Fortney 2005), and thus constraining for such a pressure range is very important for the understanding of Jovian cloud dynamics.

New insights about the Jovian atmosphere will be extended to the research of exo-planets because the Jovian atmosphere is the base for atmospheric modeling of exo-planets (Seager & Deming 2010). In particular, since our new technique will provide us the transmission spectrum of Jupiter projected on the Galilean satellites eclipsed in the Jovian shadow as screens, such a transmission spectrum of Jupiter will be applied to the modeling of the atmosphere of extrasolar giant planets by transit spectroscopy (Brown 2001). Transit spectra of 11 exo-planets have been obtained to date (Swain et al. 2014), and in these spectra, methane (Swain et al. 2008) and haze (Pont et al. 2008) were first detected in the atmosphere of the hot-Jupiter HD189733b by near-infrared transit spectroscopies. For example, haze abundance is important to estimate the size of the exo-planet from transit observations (de Kok & Stam 2012). Thus, Jovian transmission spectrum, including methane absorption and haze forward-scattering like our data, will be a standard of modeling and comparison for the characterization of these transit spectra of exo-planets.