The PEPSI Exoplanet Transit Survey (PETS) I:
Investigating the presence of a silicate atmosphere on the super-Earth 55 Cnc e

The detection of atmospheric fingerprints from terrestrial planets outside the Solar System is an important aim of exoplanet science. A suitable method for this is to analyze transmitted stellar light through the atmosphere of an exoplanet, a method called transmission spectroscopy (Seager & Sasselov, 2000). The comparison of the atmospheric fingerprints of the solar system planets to those worlds, especially those with similar properties to Earth (see e.g. Grenfell 2017; Keles et al. 2018; Wunderlich et al. 2021), will help to increase our knowledge about their evolution.

Terrestrial exoplanets could possess different atmospheres, such as a primordial atmosphere, being rich in hydrogen and helium, or a secondary atmosphere with a larger mean molecular weight and absence of hydrogen and helium (Kite & Barnett, 2020). The presence of a secondary type atmosphere on an exoplanet can be attributed to the evolution of a primordial atmosphere. Under the absence of a magnetic field, there is a possibility that a planet in a close-in orbit can lose its hydrogen-rich primordial atmosphere and remain with a rocky surface. Strong stellar irradiation can sputter particles from the exposed surface (Schaefer & Fegley, 2009) and can build up a partly ionized silicate atmosphere or exosphere (Ridden-Harper et al., 2016; Vidotto et al., 2018). Such silicate-vapor atmospheres might be built up also by outgassing processes from molten magma on the surface (Miguel et al., 2011; Ito et al., 2015) and can contain species such as Na , K , Fe , Si , SiO , O and O₂, and probably also hydrogen due to resupply processes from magma outgassing (Kite et al., 2019).

Among the close-in terrestrial exoplanets detected to date, the super-Earth 55 Cnc e (Demory et al., 2011; Winn et al., 2011), with M_P = 8.0 $\pm$ 0.3 M_$\oplus$ and R_P = 1.88 $\pm$ 0.03 R_$\oplus$ (Bourrier et al., 2018), is one of the most studied targets, due to its short orbital period of P $\approx$ 0.66 days and bright host star with a V_mag $\approx$ 6. However, the atmospheric conditions for this planet are still under discussion, due to the applicability of different interior composition scenarios estimated from the observed planetary radius and mass.

Under the assumption of a solid interior mainly composed of iron, carbon, and silicates, the planet might have no thick atmospheric envelope, i.e. exposes mostly a rocky surface (Madhusudhan, 2012), enabling the evolution of silicate-species building up a thin envelope around the planet. On the other hand, considering a less solid rocky interior, the presence of a thick atmospheric envelope is necessary (Bourrier et al., 2018), e.g. an envelope rich in water (Demory et al., 2011; Gillon et al., 2012) or in carbon dioxide and molecular nitrogen (Angelo & Hu, 2017). A primordial gaseous envelope dominated by hydrogen-helium is mainly ruled out as it may evaporate in short timescales, which is also supported by the absence of hydrogen Ly$\alpha$ (Ehrenreich et al., 2012) and helium absorption (Zhang et al., 2021).

Different investigations made efforts to confirm one of those scenarios, by searching for atomic and molecular features applying moderate and high-resolution transmission spectroscopy using ground-based facilities:

Assuming the presence of a rocky surface, Ridden-Harper et al. (2016) investigated if the strong stellar irradiation ejected atoms from the surface by sputtering to the exosphere. In their work, a tentative absorption relative to the stellar spectrum of $\sim$2.3 × 10^-3 in the Na-D-lines at a 3-$\sigma$ level and $\sim$7 × 10^-2 in the ionized CaII h&k lines at a 4.1-$\sigma$ level is shown. However, the authors did not claim detection due to low significance of the Na absorption and variable Ca⁺ signal. Another study on the rocky surface scenario for 55 Cnc e is presented by Tabernero et al. (2020) investigating the absorption in the Na-D-lines and H$\alpha$ line. The authors showed an upper limit of (3.4 $\pm$ 0.4) × 10^-4 to the presence of the Na-D-lines and (7 $\pm$ 1) × 10^-4 to the presence of the H$\alpha$ line, not detecting absorption of these species. The non-detection of hydrogen is in agreement with the findings shown by Ehrenreich et al. (2012), confirming the absence of a primordial atmosphere. Another study that supports this result is demonstrated by Zhang et al. (2021), who did not find evidence of helium absorption on this planet.

Furthermore, different studies investigated molecular absorption features within the atmosphere of 55 Cnc e studying the presence of a secondary atmosphere. For instance, Esteves et al. (2017) and Jindal et al. (2020) investigated the water absorption in the optical wavelength range, not detecting water absorption signature. In addition, Jindal et al. (2020) searched for the absorption signature of TiO also not finding any evidence for this molecule. Note that a water-rich steam atmosphere would dissociate at high altitudes, enriching the upper atmosphere with hydrogen (Bourrier et al., 2018). Thus, the non-detection of water is in agreement with the non-detection of hydrogen (Ehrenreich et al., 2012), favoring a scenario with a dry atmosphere (Bourrier et al., 2018). Further molecular features were investigated by Deibert et al. (2021). The authors searched for absorption features from CO, CO₂, H₂O, HCN, NH₃ and C₂H₂. This study did not detect these features, but they were able to put upper volume mixing ratio limits on the last three of them.

Overall, the aforementioned studies remark that the results demonstrated are still consistent with a heavyweight atmosphere scenario as well as a rocky surface scenario, further to the possible presence of clouds which might mask absorption signature. This is a valid assumption for 55 Cnc e where the formation of mineral clouds is possible (Mahapatra et al., 2017). Nevertheless, to date, no significant absorption signature for this planet, using high-resolution transit observations, has been reported.

In this study, we make use of one high-resolution transit observation of 55 Cnc e using a larger telescope than in previous studies to search for absorption by various atmospheric silicate species. This paper is structured as follows: In Section 2 the observation and data reduction process is described. Section 3 explains the methodology of this paper. The results are presented in Section 4 and discussed and concluded in Section 5.

In this work, we search for the presence of atomic species in the high spectral resolution transmission spectrum of the super-Earth 55 Cnc e. Due to the presence of strong telluric contamination of mostly saturated oxygen and water lines within the observed wavelength range, we exclude the wavelength ranges 7580Å - 7680Å, 8120Å - 8350Å and 8940Å - 9007Å from the analysis. We employ the pixel-by-pixel approach (see Section 3.2) using a second-order polynomial function to create the combined transmission spectrum of 55 Cnc e. Here, we fit the normalized and wavelength calibrated flux value at a certain wavelength position over time and divide the data by the fit. We do this for all flux values at each wavelength position. Within the combined transmission spectrum we derive this way, we investigate the absorption by various neutral atomic and ionized species on 55 Cnc e, assumed to originate from a silicate-vapor atmosphere. In order to be able to make a preselection of the species that might build up the silicate atmosphere, we assume that the crust of 55 Cnc e contains similar species as the Earth’s crust (Lide, 2008). Table 1 shows the species with mixing ratios larger than 0.01% in the Earth’s crust composition. We inspect the transmission spectrum around the Mg i-lines at 5167.3Å, 5172.7Å and 5183.6Å, the K i- line at 7699.0Å , the H$\beta$-line at 4861.3Å and the Ca ii- lines at 8498.0Å, 8542.0Å and 8662.1Å. Furthermore, we calculate synthetic template spectra (see Section 3.1) and compare these with the combined transmission spectrum via the cross-correlation method where a large number of absorption lines are co-added (Snellen et al., 2010; Brogi et al., 2016; Hoeijmakers et al., 2018). We do this within a velocity range of $\pm$200 km/s in steps of 1 km/s and use the PyAstronomy¹¹1https://github.com/sczesla/PyAstronomy tool pyasl.crosscorrRV, expecting an upward peak at 0 km/s velocity for absorption. To calculate the significance of the detection, we divide the cross-correlation signal by the standard deviation of it excluding $\pm$20 km/s around the 0 km/s region (Hoeijmakers et al., 2018). Finally, we determine the upper limit of the atmospheric extension R_ext for each species (see Section 3.3). Note, that we do not correct for the effects of centre-to-limb variation (see e.g. Yan et al. 2017) and the Rossiter–McLaughlin effect (see e.g. Borsa et al. 2021) which are insignificant for this observation. The following provides further information regarding the data analysis.

Figure 3 shows the combined transmission spectrum around strong opacity bearing species. Here, the standard deviation at $\pm$1Å around the line core is 6.05 $\times$ 10^-4 at the H$\beta$-line, 2.78 $\times$ 10^-4 at the K i-line, 6.49 $\times$ 10^-4 at the Mg i-lines and 5.09 $\times$ 10^-4 at the Ca ii-lines (see Table 2). Figure 4 shows the color-coded 2D map of the residual spectra at different orbital phases for these species. None reveals evidence of absorption.

Figures 6- 9 show the cross-correlation approach searching for different atmospheric species for the scenario with the atmospheric temperature of T = 2500 K. From left to right in landscape format, the first panels show the template spectrum, the second panels the cross-correlation signal, the third panels the colour coded cross-correlation signal for the transmission spectra at different orbital phases in the velocity range and the fourth panel the colour coded K_P - RV maps. The K_P - RV maps are calculated $\pm$100 km/s around the expected velocity-semi amplitude of the planet and are used to verify that a possible absorption signal arises only at the expected K_P value and is thus of planetary origin. The inspection of the cross-correlation signals shows no evidence of absorption for any of the investigated species. The scatter remains mainly below $\pm$3 standard deviations. Also, no evidence of significant absorption is visible in the third panels or visible at different K_P positions in the fourth panels.

The fifth panel shows the detection thresholds for different R_ext values applied for the injection recovery up to 3 R_P, with the blue dot showing the significance for the injection of the unscaled template calculated according to the planetary properties as described in Section 3.1. Due to the high-mean molecular weight of the atmosphere, the extension of the atmosphere is mainly below 1.4 R_P in the unscaled templates for all species investigated, not detectable at a 3$\sigma$ confidence level within the observation available. We are able to put an upper limit of the atmospheric extension for the investigated species by injecting scaled templates into the data and recovering them. Figure 5 shows the atmospheric extension R_ext limits which we would able to probe for the investigated species for the T = 2500 K scenario (red dots) and the T = 5000 K scenario (blue squares) at the 3$\sigma$ confidence level. The dashed line marks the planetary surface at 1 $\times$ R_P. The upper limit of the atmospheric extension shows a decrease for most species going from T = 2500 K scenario to the T = 5000 K scenario, mainly due to the variation in the number of absorption lines, but depends also on the quality of the combined transmission spectrum at the regions compared to the synthetic templates. Table 3 shows the derived limits for the investigated species at a 3$\sigma$ confidence level, with a median value of R_ext = 1.9 $\times$ R_P for the T = 2500 K scenario and R_ext = 1.7 $\times$ R_P for the T = 5000 K scenario, ruling out absorption from an atmospheric envelope which is widely extended for most of the species.

The search for absorption by silicate species in the transmission spectrum of 55 Cnc e, for instance released by a magma ocean, did not show an absorption for any of the investigated species. To account for different conditions in the upper atmosphere of 55 Cnc e, we repeated the cross-correlation analysis with template spectra calculated for T = 5000 K i.e. $\sim$2 $\times$ T_eq, yielding the same non-detection of any absorption signature in all cases.

Our non-detection of H$\beta$ agrees with previous investigations not finding evidence of hydrogen absorption for this planet (Ehrenreich et al., 2012; Tabernero et al., 2020). Other absorption signatures from species with strong opacities previously resolved on gaseous exoplanets such as Mg i (Cauley et al., 2019), K i (Keles et al., 2019) or Ca ii (Yan et al., 2020) are not detected either for 55 Cnc e.

For short period exoplanets, high escape rates of Na i, Ca ii and Mg ii can be expected to build up large planetary tails on hot rocky exoplanets (Mura et al., 2011). In this study, we derived limits for the atmospheric extension related to the absorption of major species (see Table 3), including the major abundant species in the Earth’s crust such as Al, Fe, Ca, Na, Mg and K (see Table 1), which are far below the Roche lobe radius of R_Roche = 5.35 $\times$ R_$\oplus$ (Ridden-Harper et al., 2016), i.e. R_Roche $\approx$ 2.85 $\times$ R_P, showing that this might be not the case for 55 Cnc e.

We compare our results to the findings by Ridden-Harper et al. (2016) who showed a possible 3-$\sigma$ sodium signature at the Na d lines and a variable 4.1-$\sigma$ Ca⁺ signature at the Ca h&k lines on 55 Cnc e. As Ridden-Harper et al. (2016) uses a larger planetary radius which bases on Gillon et al. (2012), we compare the inferred probed altitudes in terms of R_$\oplus$. The Na d signal from Ridden-Harper et al. (2016) corresponds to R_P = 5 $\times$ R_$\oplus$. To be able to compare our findings,we scale the Na- template so that the Na d lines in the templates reach this altitude and repeat the analysis. Note, that we have no access to the resonance lines in our observations, but the scaling affects all lines within the wavelength range accessible. For Na d lines reaching 5 $\times$ R_$\oplus$ in the template, the injection recovery would yield a 5.8-$\sigma$ detection. Therefore, we can rule out the presence of sodium at such extended altitudes. In our work, the inferred 3-$\sigma$ limit for the Na investigation shows an upper limit of R_ext = 1.73 $\times$ R_P, which would correspond to Na d lines reaching R_ext = 1.99 $\times$ R_P and thus R_P = 3.75 $\times$ R_$\oplus$ compared to the R_P = 5 $\times$ R_$\oplus$ presented by Ridden-Harper et al. (2016). We apply the same comparison for the Ca⁺ lines. The Ca h&k lines reach around R_P = 25 $\times$ R_$\oplus$ in Ridden-Harper et al. (2016), corresponding to R_ext $\approx$ 13.30 $\times$ R_P. We scale the ionized calcium- template so that the Ca h&k lines in the templates reach this altitude and repeat the analysis. Note, that we have also no access to these resonance lines in our observations. For such an atmospheric extension, we would yield a > 64-$\sigma$ detection, which is not the case. A 4.1-$\sigma$ detection of ionized calcium would correspond to an upper limit of around R_ext = 2.1 $\times$ R_P in this work, which would correspond to Ca h&k lines reaching R_ext = 2.39 $\times$ R_P and thus R_P = 4.5 $\times$ R_$\oplus$, compared to the R_P = 25 $\times$ R_$\oplus$ presented by Ridden-Harper et al. (2016). Thus, within the direct comparison to Ridden-Harper et al. (2016) regarding the Na- and Ca⁺- absorption, we can only put a slightly lower absorption limit regarding the presence of sodium, but more than five times lower limit regarding the presence of the ionized calcium in the exosphere of 55 Cnc e, probably due to the larger S/N values achieved by the observation and the cross-correlation analysis, although Ridden-Harper et al. (2016) have access to stronger resonance absorption lines.

Note, that Ridden-Harper et al. (2016) did not claim detection and reported the variability of the Ca h&k absorption, which might be due to stellar activity. Probably also induced by stellar variability, a broad downward showing feature is visible in the cross-correlation analysis for the ionized Ca ii lines (see Figure 6) at a significance of $\sim$4 $\sigma$. As the template spectrum and the transmission spectrum have the same alignment for the cross-correlation analysis, a downward peaking feature suggests emission rather than absorption. Inspecting the Ca ii- infrared triplet lines that have the major contribution within the cross-correlation analysis, in the combined transmission spectrum (Figure 3), only the Ca ii line around 8498Å shows a hint for an emission. We repeat the cross-correlation analysis for the ionized Ca ii excluding this line, which results in the decrease of the significance of the feature down to $\sim$3 $\sigma$. Therefore, this strengthens the possibility that the emission feature does not arise from the planet or from an ionized calcium-rich envelope, but is rather due to variability in the stellar calcium lines, also strengthened by the fact that the emission is not following the atmospheric trace (see the third panel in Figure 6).

The atmospheric conditions on 55 Cnc e are still puzzling: is there a detectable atmosphere on this planet? Possible scenarios include a purely rocky surface, a rocky surface with a magma ocean on top releasing species and building up a thin silicate-vapor envelope, or a atmosphere e.g. with a water envelope or a N₂ or CO₂ rich atmosphere (Angelo & Hu, 2017) with a surface pressure similar to Earth. The non-detection of water (Esteves et al., 2017; Jindal et al., 2020) with the non-detection of light elements (Ehrenreich et al., 2020; Zhang et al., 2021) seems to make the water envelope scenario unlikely. A detectable thin silicate-vapor atmosphere scenario can not be strengthened either by Tabernero et al. (2020) who did not find evidence for Na i absorption or by this work not finding evidence of absorption for a variety of species. Overall, our work provide indications that favor the heavyweight atmospheric scenario. This would be in agreement with the shift detection of the hot spot east of the substellar point on 55 Cnc e detected by Demory et al. (2016). Such a shift is usually driven by atmospheric circulation processes, including a significant amount of atmospheric mass (Showman & Polvani, 2011; Miller-Ricci Kempton & Rauscher, 2012; Keles, 2021). Although we cannot rule out a heavy silicate-atmosphere existing below the presented altitude limits inferred in this work, with a median value of R_ext = 1.9 $\times$ R_P, or a deck of mineral clouds (Mahapatra et al., 2017) that hide absorption signatures, this study showed that very high S/N data can deliver informative upper limits on the extension of the possibly existing silicate atmospheres probing deeply the atmospheres of terrestrial exoplanets.

This survey is a collaborative effort from AIP, MPIA, INAF, University of Arizona, Ohio State University and the LBT.
EK is thankful for the support and granting the LBT time for this observation by the Rat Deutscher Sternwarten (RDS).
XA is grateful for the financial support from the Potsdam Graduate School (PoGS) in the form of a doctoral scholarship.
KM: This research was supported by the Excellence Cluster ORIGINS which is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy - EXC-2094 - 390783311.
KP and LK acknowledge support from the Leibniz-Gemeinschaft under project number P67/2018.

The data underlying this article will be shared on reasonable request to the corresponding author.