THE ULTRAVIOLET RADIATION ENVIRONMENT AROUND M DWARF EXOPLANET HOST STARS^*

The stellar Lyα emission line dominates the UV output of M dwarfs, containing approximately as much energy as the rest of the FUV+NUV spectrum combined (France et al. 2012a). Therefore, measurements of Lyα emission provide an important constraint on the source term for calculations of the heating and photochemistry of exoplanetary atmospheres (Linsky et al. 2012a). Due to resonant scattering of neutral hydrogen and deuterium in the interstellar medium (ISM), the intrinsic Lyα radiation field cannot be directly measured, even in the nearest stars (log₁₀ N(H i) ≲ 18.5). In order to produce an accurate estimate of the local Lyα flux incident on the planets around our target stars, the Lyα profiles observed by STIS must be reconstituted. We do this using a technique that simultaneously fits the observed Lyα emission line profile and the ISM component, using the MPFIT routine to minimize χ² between the fit and data (Markwardt 2009).

The intrinsic Lyα emission line is approximated as a two-component Gaussian, as suggested for transition region emission lines from late-type stars by Wood et al. (1997). This approximation is further justified by the reconstructed Lyα profiles for M dwarfs, which show approximately Gaussian line shapes and little evidence for self-reversal compared to F, G, and K dwarfs (Wood et al. 2005). The Lyα emission components are characterized by an amplitude, FWHM, and velocity centroid. Interstellar absorption is parameterized by the column density of neutral hydrogen (N(H i)), Doppler b parameter, D/H ratio, and the velocity of the interstellar absorbers. We assumed a fixed D/H ratio (D/H = 1.5 × 10⁻⁵; Linsky et al. 1995, 2006). Degeneracy between the two emission components makes errors difficult to define for the individual fit parameters. We estimate the uncertainty on the integrated Lyα fluxes from our technique by comparing the reconstructed Lyα flux for a range of initial guesses on the parameters. The uncertainty on the integrated intrinsic flux is ≈10%–20% for the stars observed with E140M (GJ 832 and GJ 667C) and ≈15%–30% for the stars observed with G140M (GJ 581, GJ 876, GJ 436). In order to test our methodology, we fitted the STIS E140M Lyα profile of AU Mic. Our total integrated flux agrees with that presented in Wood et al. (2005) to ≈5%, with an identical derivation of the interstellar atomic hydrogen column on the line of sight (log₁₀ N(H i) = 18.36 ± 0.01).

Employing our iterative least-squares reconstruction technique, we also derived the intrinsic Lyα spectrum of HD 189733. Using the 2010 April 6 STIS G140M spectra (Lecavelier des Etangs et al. 2012), we find an intrinsic flux of F(Lyα) = 7.5 × 10⁻¹³ erg cm⁻² s⁻¹ and an interstellar hydrogen column density log₁₀ N(H i) = 18.45. We also refitted the intrinsic Lyα emission from AD Leo, finding F(Lyα) = 7.5 × 10⁻¹² erg cm⁻² s⁻¹ and an interstellar hydrogen column density log₁₀ N(H i) = 18.47 ± 0.02. This value agrees with the Lyα reconstruction from Wood et al. (2005) to ≈30%, which gives us confidence in the iterative technique, even for sightlines with multiple interstellar velocity components.

We present the individual Lyα spectra in Figure 2, and narrow/broad emission component ratios (F_narrow/F_broad), spectral line widths, H i column densities, and Doppler b-values in Table 2. We find F_narrow/F_broad to be in the range 5–12 for the MUSCLES stars, again independent of the resolution of the observations on which the reconstructions are based. For comparison, we find F_narrow/F_broad ∼3 for the more active M dwarf (AD Leo) and more massive star (HD 189733). Reconstructed Lyα fluxes are given in Tables 3, 4, and 5. GJ 1214 has no Lyα emission detected, and we discuss this in greater detail in Section 5.1.1.

Zoom In Zoom Out Reset image size

Table 2. Stellar Lyα Emission Line Widths and Interstellar Atomic Hydrogen^a

Target	Fnarrow/Fbroad	FWHMnarrow	FWHMbroad	log10 N(H i)	$b_{{\rm H\,\scriptsize{I}}}$
		(km s−1)	(km s−1)	(cm−2)	(km s−1)
GJ 581b	8.4	110	228	18.35 ± 0.06	10.1
GJ 876b	5.0	132	303	18.06 ± 0.03	8.0
GJ 436b	11.8	109	294	18.19 ± 0.04	7.4
GJ 832c	7.6	96	163	18.47 ± 0.02	9.7
GJ 667Cc	8.9	85	233	18.07 ± 0.03	12.7
AD Leoc	2.9	166	412	18.47 ± 0.01	9.0
HD 189733b,d	3.5	170	480	18.45	10

Notes. ^a Lyα profiles reconstructed using the iterative least-squares technique described in Section 4.1. ^bSTIS G140M observations. ^cSTIS E140M observations. ^dHD 189733 reconstruction failed to converge, the presented profile parameters provide a fit to the observed line profile that is qualitatively similar to the formal fits for the M dwarf spectra.

Download table as:  ASCIITypeset image

In our initial study of GJ 876, we described the large chromospheric/transition region flare observed in several UV emission lines during our COS observations (France et al. 2012a). We have carried out a similar analysis on the entire MUSCLES sample to constrain the heretofore unknown level of UV variability in optically quiet M dwarf exoplanet host stars. Light curves in several chromospheric and transition region lines were extracted from the calibrated three-dimensional data by exploiting the time-tag capability of the COS microchannel plate detector (France et al. 2010a). We extract a [λ_i,y_i,t_i] photon list (where λ is the wavelength of the photon, y is the cross-dispersion location, and t is the photon arrival time) from each exposure i and combine these to create a master [λ,y,t] photon list. The total number of counts in a [Δλ,Δy] box is integrated over a time step Δt. We use a flux-dependent time step of Δt = 40–120 s for the MUSCLES targets. The instrument background level is computed in a similar manner, with the background integrated over the same wavelength interval as the emission lines, but physically offset below the science region in the cross-dispersion direction. The signal-to-noise ratio (S/N) was not high enough to temporally resolve the GJ 1214 data.

Chromospheric and transition region emission lines are observed in all of the MUSCLES spectra, suggesting that many M dwarf exoplanet host stars have UV-active atmospheres (Figure 1). The chromosphere is observed in H i Lyα, C ii λλ1334, 1335, Al ii λ1671, Fe ii multiplets near 2400 and 2600 Å, and Mg ii λλ2796, 2803; lines that are formed from ∼1–3 × 10⁴ K. The transition region is traced by lines with formation temperatures from ∼0.6–1.6 × 10⁵ K, including the C iii 1175 multiplet, Si iii λ1206, O v λ1218, N v λλ1239, 1243, Si iv λλ1394, 1403, C iv λλ1548, 1550, and He ii λ1640. C ii and C iv line fluxes for GJ 581, GJ 876, GJ 436, and GJ 832 have been published in Linsky et al. (2012a), and a comprehensive set of stellar emission line measurements will be presented in a future work. In Table 5, we present the reconstructed Lyα fluxes and the observed fluxes of C iv and Mg ii.

Figure 3 displays the C iv profiles for GJ 581, GJ 876, GJ 436, and GJ 832 scaled to the flux level at 0.16 AU from the parent star, a value near the center of the HZ for the MUSCLES targets.⁸ While we conclude that most M dwarf exoplanet hosts have a non-zero local UV radiation field, the amplitude of this field is quite variable. The HZ C iv fluxes from these four stars varies by a factor of 10 from the weakest (GJ 581) to the strongest (GJ 832) C iv emitter. This is particularly interesting in light of the established correlation between X-ray and C iv luminosity for M dwarf emission line (dMe) flare stars (Byrne & Doyle 1989). With improved atmospheric models for M dwarfs or simultaneous UV and X-ray observations of a larger sample of M dwarf planet hosts, the C iv flux in the HZ of low-mass stars could be calibrated as a proxy for the EUV and soft X-ray emission incident on the planetary atmospheres (Linsky et al. 2012a). C iv-based EUV irradiance estimates could provide a complementary constraint to those derived from coronal X-ray measurements (Sanz-Forcada et al. 2011).

Zoom In Zoom Out Reset image size

Lyα is the brightest line in the UV spectrum of low-mass stars, although resonant scattering in the ISM makes direct line profiles inaccessible, even for the nearest stars (Wood et al. 2000, 2005). As described above, Lyα reconstruction can be a powerful technique for determining the intrinsic stellar Lyα line flux, but this technique is not possible for more distant targets where the ISM completely removes stellar Lyα photons from our line of sight (N(H i) ≳ 10²⁰ cm⁻²). In this case, one requires a proxy formed in a similar region of the chromosphere to estimate the local Lyα emission flux. For our limited sample, we have found that Lyα and Mg ii are well correlated (see also Linsky et al. 2012a), and we present a scaling relation to infer the Lyα flux from an Mg ii observation.

Figure 4 shows the reconstructed Lyα flux versus the observed Mg ii flux, scaled to 0.16 AU from the host star for GJ 581, GJ 876, GJ 436, GJ 832, and GJ 667C (black filled circles). A linear fit to the correlation (F(Lyα) = a + bF(Mg ii), where a = −2.3 ± 2.2 and b = +14.8 ± 1.2), is also displayed in Figure 4. The observed F(Lyα)/F(Mg ii) ratio is the slope of this fit. Chromospheric Mg ii emission lines of G and K dwarfs are known to show self-reversed profiles due to non-LTE effects in their upper atmospheres (e.g., Linsky & Wood 1996; Wood & Linsky 1998; Redfield & Linsky 2002); however, the situation is less clear for M dwarfs. For example, Redfield et al. (2002) present GHRS spectra of AU Mic, where no appreciable self-reversal of the reconstructed Mg ii profile is observed. We therefore assume a simple Gaussian line shape for the M dwarf Mg ii emission lines. Interstellar Mg ii absorption will produce the appearance of a self-reversed emission line when the stellar velocity is close to the velocity of the Mg⁺-bearing interstellar cloud (e.g., Wood et al. 2000); therefore a correction for interstellar Mg ii absorption needs to be made in order to give an unbiased estimate of the F(Lyα)/F(Mg ii) ratio for the MUSCLES stars.

Zoom In Zoom Out Reset image size

We cannot analyze the interstellar Mg ii absorption in detail for most of our targets because the stellar and interstellar features are unresolved in STIS G230L observations. However, utilizing our STIS E230H high-resolution (Δv < 3 km s⁻¹) spectra of GJ 832 and GJ 667C, the correction for interstellar Mg ii absorption can be estimated. The profile of the Mg ii k line (λ_o = 2795.528 Å) of GJ 832 in shown in Figure 5. For GJ 832 and GJ 667C, we find that the interstellar absorber removes 30%–35% of the intrinsic Gaussian emission line flux. Therefore, we estimate that the scaling of the ISM-corrected Mg ii emission line flux to the intrinsic Lyα emission line flux is ≈10. Factoring in uncertainties on the interstellar correction and the Lyα reconstruction, we estimate that this scaling relation is good to ∼30%, or the ISM-corrected intrinsic F(Lyα)/F(Mg ii) ratio for weakly active M dwarf exoplanet host stars is 10 ± 3. Linsky et al. (2012a) find that the F(Lyα)/F(Mg ii) ratios for the more active M dwarfs AU Mic and AD Leo are 2.5 and 4.0, respectively. This suggests that the ratio may decrease with activity.

Zoom In Zoom Out Reset image size

5.1.1. Non-detection of Lyα in GJ 1214

GJ 1214 is the coolest (M6V, T_eff 2949 ± 30 K; Kundurthy et al. 2011) and most distant source in the MUSCLES pilot study, and the only star where H i Lyα is not detected. In the top panel of Figure 6, we display the full UV spectrum of GJ 1214. The only clearly detected stellar emission lines are those of C iv and Mg ii. In the lower panel, we focus on the immediate spectral region around the lines of interest. The upper limit from the co-added STIS G140M spectra of GJ 1214 is F(Lyα) < 2.4 × 10⁻¹⁵ erg cm⁻² s⁻¹. The short-wavelength component of the C iv doublet is shown in red, and we show a possible detection of N v in blue. Taking the observed Mg ii flux from GJ 1214 (2.21 (±0.15) × 10⁻¹⁵ erg cm⁻² s⁻¹), we can use the F(Lyα)/F(Mg ii) relation described above to compare the expected flux from Lyα with the upper limit (red circle and red star in Figure 4). The upper limit on the GJ 1214 Lyα flux is an order of magnitude below the predicted point based on the observed Mg ii flux. Note however that this relation is for reconstructed Lyα emission lines and we have no information on which to base a reconstruction in GJ 1214.

Zoom In Zoom Out Reset image size

The presence of the UV metal line emission demonstrates that GJ 1214 maintains some basal UV flux level. However, the non-detection of Lyα is surprising in light of the results presented above. We suggest three possible explanations that might account for the non-detection. First, the 2012 COS and STIS G230L observations may have occurred during a period of stellar activity, whereas the star may have been UV-quiet during the 2011 STIS G140M observations. This type of "binary" UV activity behavior has been observed previously in the M8 flare star VB 10 (Linsky et al. 1995). Second, if the GJ 1214 Lyα emission line was significantly narrower than the Lyα lines in the other M dwarfs, then it might be possible that the somewhat longer interstellar sightline removed the signal. However, as shown by Wood et al. (2005), neutral interstellar hydrogen columns rarely exceed log₁₀ N(H i) = 18.5 cm⁻² for stars nearer than 50 pc. Therefore, this second option is really that the stellar Lyα emission line from GJ 1214 is unusually narrow. There is not enough information from C iv or Mg ii to make a meaningful comparison of GJ 1214 line widths to the other stars in our sample. The third, more speculative, possibility is that for the low effective temperature of GJ 1214, a significant fraction of the available neutral hydrogen in the stellar atmosphere is locked up in molecules (mainly H₂). As we will discuss in Section 5.3, we detect H₂ emission in all of the other M dwarfs observed with COS. We do not detect H₂ fluorescence in GJ 1214, but this could be the consequence of a feedback process driven by increasing molecular fraction. Lyα is the energy source for the observed fluorescence, so as fewer atoms are available to produce Lyα, the H₂ fluorescence process will become unobservable in FUV observations even if more molecules are present in the stellar atmosphere.

In earlier work describing the UV spectrum of GJ 876, it was shown that the FUV/NUV flux ratio is >10³ times the solar value (France et al. 2012a). This is a consequence of an enhanced Lyα line strength and a much lower NUV flux owing to the cooler effective temperature of GJ 876. In Figure 7, we show that this is a generic property of M dwarfs. Here, we define the FUV flux, F(FUV), as the total stellar flux integrated over the 1160–1690 Å bandpass, excluding Lyα. Therefore, the total FUV emission is ΣFUV = (F(Lyα) + F(FUV)), the total FUV/NUV ratio is

$$\begin{equation} \Sigma {\rm FUV/NUV} = (F({\rm Ly}\alpha) + F({\rm FUV}))/F({\rm NUV}), \end{equation} \tag{ 1 }$$

and the Lyα fraction is defined as

$$\begin{equation} f({\rm Ly}\alpha) = F({\rm Ly}\alpha) / (F({\rm Ly}\alpha)+ F({\rm FUV}) + F({\rm NUV})), \end{equation} \tag{ 2 }$$

where F(Lyα) is integrated over 1210–1222 Å and F(NUV) is integrated over 2300–3050 Å.

Zoom In Zoom Out Reset image size

The observed Lyα, FUV, and NUV fluxes are presented in Table 3. We put these values in context in Table 4 by presenting the Lyα, FUV, and NUV luminosities as the fraction of the bolometric luminosities. We calculate the bolometric luminosity for each star by creating a continuous spectrum from 0.115 to 2.5 μm, using the appropriate M dwarf model atmosphere (Pickles 1998) to cover the optical and IR portions of the spectral energy distribution. The model atmosphere flux is scaled to the STIS G230L data from 3050–3080 Å. The use of the Pickles models will underestimate the bolometric luminosity somewhat (by not including the entire Rayleigh–Jeans tail into the mid-IR); however, we expect that this effect is similar to or smaller than the uncertainty introduced by scaling to the NUV data, where the relative contribution of the photosphere is not clear in most cases.

ΣFUV/NUV is in the range ≈ 1–3 for all M dwarfs, except GJ 1214. Mostly due to the lack of Lyα emission (Section 5.1.1), ΣFUV/NUV is in the range 0.50–0.62 for GJ 1214. While the ΣFUV/NUV ratio is lower for GJ 1214 than for the rest of the M dwarfs in our sample, it is still significantly higher than for earlier type stars. For HD 189733, this ratio has fallen to ∼10⁻², and it is < 10⁻³ for the Sun. Similarly, f(Lyα) spans 37%–75% for most M dwarfs, is < 7.4% for GJ 1214, is ≈0.7% for HD 189733, and is ≈0.04% for the Sun. AD Leo has an absolute NUV flux that is 4–40 times larger than for the less active M dwarfs in this survey, and the ΣFUV/NUV ratio could decrease dramatically during flares when strong Balmer continuum emission dominates the spectrum (Hawley & Pettersen 1991). The implications of the large FUV/NUV flux ratio in M dwarfs are discussed in more detail below (Section 7.1).

We observe fluorescent H₂ emission lines in GJ 581, GJ 876, GJ 436, and GJ 832. These features are photoexcited ("pumped") by Lyα photons out of the v = 2 level of the ground electronic state (Shull 1978). Under the assumption of thermalized rovibrational populations, this process is a signpost for 2000–4000 K molecular gas. We detect between four and eight fluorescent emission lines in each of these targets, pumped by Lyα through the (1–2)R(6) λ1215.73 Å and (1–2)P(5) λ1216.07 Å transitions. The S/N of any of the individual lines is not sufficiently high for spectral fitting, so we co-added four to six of the brightest lines.⁹ The co-added line profiles were normalized to unity and fitted by assuming a Gaussian emission line shape that has been convolved with the 1500 Å HST+COS line spread function (Kriss 2011). The co-added fluorescence lines and fits are shown in Figure 8. The H₂ emission lines are consistent with the stellar radial velocities to within half of the 15 km s⁻¹ wavelength solution uncertainty of the COS G160M mode, [v_H2(581) = −5.5 km s⁻¹, v_H2(876) = +9.0 km s⁻¹, v_H2(436) = +17.5 km s⁻¹, v_H2(832) = +19.7 km s⁻¹] compared with [v_rad(581) = −9.5 km s⁻¹, v_rad(876) = +8.7 km s⁻¹, v_rad(436) = +10 km s⁻¹, v_rad(832) = +18.0 km s⁻¹]. Except in the case of GJ 581 (the lowest S/N H₂ spectrum), the H₂ emission features are spectrally unresolved. In this subsection, we describe three possible physical origins for the observed fluorescence: (1) the stellar photospheres, (2) the atmospheres of the planets orbiting these stars, or (3) circumstellar gas disks populated by atmospheric mass loss from the planets in these systems.

Zoom In Zoom Out Reset image size

Lyα-pumped H₂ emission lines were first detected in sounding rocket spectra of sunspots (Jordan et al. 1977), and have been modeled assuming a thermalized H₂ population at a temperature of 3200 K (Shull 1978). H₂ fluorescence lines are also powerful diagnostics of the molecular surface layers of protoplanetary disks over a similar range of temperatures (Herczeg et al. 2004; France et al. 2012b). These lines are not clearly detected in previous STIS observations of M dwarfs without planets (AD Leo, Proxima Cen, EV Lac, and AU Mic¹⁰), although this could be related to much larger instrumental backgrounds associated with the STIS echelle modes. In any event, the 3200 K sunspot temperature is characteristic of the photospheric effective temperature of the MUSCLES stars, and simple energy budget calculations (e.g., France et al. 2012a) suggest that the fluorescence may be consistent with an origin in a hot molecular stellar surface layer.

However, it is interesting that these lines are either not detected or are marginally detected with much lower H₂/atomic line ratios in the non-planet-hosting stars. H₂ is the primary constituent of gas giant planets. After H i Lyα, the electronically excited emission spectrum of H₂ dominates the FUV spectrum of giant planets (Broadfoot et al. 1979; Wolven & Feldman 1998; Gustin et al. 2010). Fluorescence driven by solar Lyβ is the main constituent of the dayglow (non-auroral illuminated disk) spectrum of Jupiter (Feldman et al. 1993) and in areas where the Jovian surface temperature is elevated (e.g., the Shoemaker–Levy 9 impact site; Wolven et al. 1997), Lyα-pumped fluorescence becomes strong.

All four of the stars showing strong H₂ fluorescence host planets of Neptune to super-Jupiter mass. Our analysis cannot rule out the possibility that the ubiquitous Lyα-pumped H₂ fluorescence observed toward M dwarf planet hosts is the result of photoexcited dayglow emission in planetary atmospheres. In light of the recent tentative detections of FUV H₂ emission from the planetary systems around HD 209458 (France et al. 2010b) and brown dwarf 2MASS J12073346−3932539 (France et al. 2010a), it is tempting to attribute an exoplanetary origin to these lines. We caution that further study is required to differentiate between these two scenarios. If the H₂ in the GJ 436 system originates in the atmosphere of GJ 436b, we would have expected to see a significant velocity shift of the fluorescent lines relative to the stellar radial velocity (v_436b = 110–120 km s⁻¹ at orbital phase ≈0.8). However, the H₂ lines are consistent with the radial velocity of the star, indicating little direct dayglow and/or auroral contribution from GJ 436b. The upper limit on the H₂ (1–7)R(3) emission line flux at +110 km s⁻¹ (assuming unresolved emission lines) is ≈6 × 10⁻¹⁷ erg cm⁻² s⁻¹ in the GJ 436 spectrum. Modeling of both the H₂ fluorescence spectrum from M dwarf photospheres and the expected UV emission spectrum from nearby M dwarf planetary systems would be extremely valuable.

A final possibility is that the observed H₂ fluorescence originates in a circumstellar gas envelope that is being replenished by atmospheric mass loss from escaping planetary atmospheres. The narrow H₂ line width in GJ 436 rules out the possibility that the fluorescing gas is confined to the semimajor axis of the planetary orbit (a = 0.03 AU for GJ 436b); however, over the ∼5 Gyr lifetimes of these systems, this material could disperse into a low-density "second generation" circumstellar disk. While detailed modeling of such a disk is beyond the scope of this work, we can make a rough estimate of the planetary mass-loss rate required to sustain the observed H₂ emission.

The balance equation of H₂ photodissociation and replenishment is β_dissn(H₂) = Φ_loss. Due to the negligible 912–1120 Å continuum flux from inactive M-type stars, we assume that the interstellar radiation field (ISRF) dominates H₂ dissociation. The photodissociation rate is β_diss = B_dissβ_o. B_diss is the dissociation fraction of molecules excited into the Lyman and Werner bands following absorption of a 912–1300 Å photon, ≈0.2 for a hot H₂ gas. β_o is the standard H₂ photoabsorption rate from the ISRF, β_o = 5 × 10⁻¹⁰ s⁻¹ (Jura 1975). We assume an H₂ column density of N(H₂) ≈ 10¹⁷ cm⁻² (France et al. 2012a) and a fiducial cloud diameter of d_H2 = 1 AU such that n(H₂) = N(H₂)/d_H2 ≈ 7 × 10³ cm⁻³. The replenishment term from the escaping atmospheres of our short-period planets is Φ_loss = $\dot{M}_{\rm loss} m_{{\rm H2}}^{-1}$ F(H₂/H i) V⁻¹_emit, where $\dot{M}_{\rm loss}$ is the planetary atmosphere mass-loss rate, F(H₂/H i) is the fraction of the total cloud in molecular form, and V_emit is the volume of the emitting area (V_emit = πr²_H2z_H2). Models of escaping hot Jupiter atmospheres find that most of the escaping material is atomic (e.g., Tremblin & Chiang 2013), so we assume a low molecular fraction F(H₂/H i) = 10⁻³. Any disk populated by an orbiting planet will most likely be geometrically thin, and we take z_H2 = 0.01 d_H2 as the vertical height. Keeping the crudeness of this approach in mind, we find that the H₂ disk can be sustained for mass-loss rates $\dot{M}_{\rm loss}\gtrsim$ 6 × 10¹⁰ g s⁻¹, well within the range of mass-loss estimates of short-period planets around G-, K-, and M-type stars (Vidal-Madjar et al. 2008; Linsky et al. 2010; Lecavelier des Etangs et al. 2012; Ehrenreich et al. 2011; Ehrenreich & Désert 2011). Therefore, the H₂ fluorescence observed in our MUSCLES sample appears compatible with either a photospheric or circumstellar origin.

Owing to the scarcity of M dwarf exoplanet host spectra previously available in the literature, authors modeling the atmospheres of Earth-like planets around M dwarfs have typically been driven to one of two extremes: assuming no UV flux (Segura et al. 2005; Rauer et al. 2011; Kaltenegger et al. 2011) or assuming the spectrum of one of the most active M dwarfs known, AD Leo (Segura et al. 2005; Wordsworth et al. 2010). In the case of no UV flux, the short-wavelength spectral cutoff is determined by the photosphere of a theoretical star with T_eff ≲ 3500 K, typically having negligible flux below 2500 Å and no chromospheric emission features. Walkowicz et al. (2008) demonstrated that even M dwarfs with modest X-ray emission and no chromospheric Hα emission display NUV spectra that are qualitatively similar to more optically active stars. Following this work, Kaltenegger et al. (2011) noted that an active UV atmosphere is most likely present on almost all M dwarfs, which is consistent with the observations presented here. Additional observations are required to understand the UV flux from M dwarfs as a function of stellar age and metallicity, for spectral types later than M6, and for a larger sample of M dwarfs without X-ray emission or Hα activity.

The MUSCLES observations show that the UV spectral behavior of four of our stars (GJ 581, GJ 876, GJ 436, and GJ 832) is similar to AD Leo in both the spectral and the temporal sense. The MUSCLES stars (except GJ 1214) showed similar a ΣFUV/NUV flux ratio and f(Lyα) to that of AD Leo, and they also displayed time variability with UV emission line amplitudes changing by factors of ≳ 2 during our brief observing snapshots.

While AD Leo appears to be a more appropriate choice for exoplanetary modeling input, a primary difference is the absolute flux of the NUV radiation field from AD Leo. AD Leo has factors of ∼4–40 stronger absolute flux levels at its effective HZ distance than do the rest of the MUSCLES targets. Furthermore, the strong Balmer continuum associated with its frequent flare activity will periodically drive the ΣFUV/NUV ratio significantly below the level (∼1–3) found for the less active M dwarfs. This could lead to significant differences between the atmospheric chemistry of a planet orbiting an AD Leo-like star and one orbiting a more typical M dwarf. It has been recently shown that the strong FUV flux and high ΣFUV/NUV flux ratio incident on a terrestrial planet orbiting in the HZ of an M dwarf can lead to the buildup of a large, abiotic atmospheric O₂ abundance through the photodissociation of CO₂ (Tian et al. 2012). The creation of O₃ via O₂ photolysis depends sensitively on the strength of the 1700–2400 Å radiation field and the balance of the ocean–atmosphere system on these planets. In some models, this process can lead to detectable levels of O₃ without the presence of an active biosphere (Domagal-Goldman et al. 2012).

Spectral observations of O₂ and O₃, along with CH₄ and CO₂, are expected to be the most important signatures of biological activity on planets with Earth-like atmospheres (Des Marais et al. 2002; Seager et al. 2009), which may be the case for such planets orbiting solar-type stars. However, the above work has shown that when realistic inputs for the UV radiation field are included, it may be possible to produce O₂ and O₃ with observable abundances on terrestrial planets in the HZ of M dwarfs without the presence of biological life. Therefore, a detailed knowledge of the stellar spectrum will be critical to interpreting the observations of potential biomarkers on these worlds when they are detected in the coming decades. Until such time that stellar models of M-type stars can reliably predict the observed UV-through-IR spectrum, a direct UV observation will be the most reliable means for determining the radiation field incident on planets orbiting M dwarfs.

In Section 6, we described the range of time variability observed in the UV chromospheric and transition region emission lines from our M dwarf exoplanet host stars. We observe emission line flux variability of 50%–500% on timescales shorter than the typical transit duration of short-period exoplanets (few hours). For comparison, the typical depth of a resolved UV transit observation is ∼5%–15% (Vidal-Madjar et al. 2003; Ben-Jaffel 2007; Linsky et al. 2010; Lecavelier des Etangs et al. 2012). The importance of stellar variability for UV transit studies of G and K dwarfs is just now being explored (Haswell et al. 2012). While an older solar-type transiting planet host star such as HD 209458 has been shown to have very steady transition region emission (observed in Si iv; Linsky et al. 2010), UV variability on younger, less-massive transiting planet hosts like HD 189733 may significantly hinder the ability to detect transits and characterize their depth and ingress/egress profiles. GJ 436 and GJ 1214 both host transiting planets (Neptune and super-Earth mass planets, respectively), and our results suggest that transit studies using UV chromospheric emission lines as background sources to search for signs of atmospheric escape may be complicated by stellar activity. NUV searches for O₃ in the atmospheres of Earth-like planets will also likely be impacted by M dwarf stellar activity, even for weakly active stars. Long-baseline observations over multiple transits will be required to mitigate the effects of stellar activity in future M dwarf exoplanet transit observations. We will consider this issue in more detail in a future work.

As part of the MUSCLES pilot program, we have started building a library of medium/high-resolution UV spectra of M dwarf exoplanet host stars. These data are available to download for use in photochemical models of Earth-like, ice giant, and gas giant planets.¹¹

The high resolution of the MUSCLES sample is important for detailed molecular emission/absorption modeling. As most of the UV flux from low-mass stars arises in discrete emission lines, molecular species in the atmospheres of HZ planets will be "selectively pumped": only species that have spectral coincidences with stellar emission lines will be subject to large energy input from the host star. H₂O, CH₄, and CO₂ in particular are highly sensitive to far-UV radiation, and provide an illustration of the utility of spectrally resolved data. CH₄ absorbs in a broad band at λ < 1500 Å, and the photoabsorption rates are relatively insensitive to the spectral resolution. CO₂ displays a banded structure over a broad absorption region from ∼1200–1600 Å, and high-resolution data show a spectral coincidence between the narrow CO₂ bands and the Si iii λ1298 + O i λ1304 multiplet and the C ii λ1335 stellar emission lines. H₂O dissociation is influenced by far-UV photons as the molecules absorb in a broad continuum from 1400–1900 Å, B–X absorption bands near 1280 Å, and sharp Rydberg transitions at λ < 1240 Å that overlap with the Lyα emission line (Yi et al. 2007). The spectral overlap between the photo-absorption cross-sections of CO₂ and H₂O and the chromospheric and transition region emission lines from an M dwarf are shown in Figure 10. Low-resolution spectra or molecular absorption cross-sections will fail to preserve the specific energy absorption in these narrow (Δλ < 10 Å) wavelength coincidences and could lead to an underestimation of the dissociation rates of important atmospheric species.

Zoom In Zoom Out Reset image size