THE MUSCLES TREASURY SURVEY. I. MOTIVATION AND OVERVIEW^*

FUV and NUV Irradiance: Photochemistry and Biosignatures—Spectral observations of O₂, O₃, 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; Kaltenegger et al. 2007; Seager et al. 2009). The chemistry of these molecules in the atmosphere of an Earth-like planet depends sensitively on the strength and shape of the UV spectrum of the host star (Segura et al. 2005). H₂O, CH₄, and CO₂ are sensitive to FUV radiation (912–1700 Å), in particular, the bright HI Lyα line, while the atmospheric oxygen chemistry is driven by a combination of FUV and NUV (1700–3200 Å) radiation (Figure 1).

Zoom In Zoom Out Reset image size

The photolysis (photodissociation) of CO₂ and H₂O by Lyα and other bright stellar chromospheric and transition region emission lines (e.g., C ii λ1335 Å and C iv λ1550 Å) can produce a buildup of O₂ on planets illuminated by strong FUV radiation fields. Once a substantial O₂ atmosphere is present, O₃ is primarily created through a multi-step reaction whereby O₂ dissociation (by 1700–2400 Å photons) is followed by the reaction O + O₂ + $\xi \;\to$ O₃ + ξ, where ξ is a reaction partner required to balance energy conservation. O₃ photolysis is then driven by NUV and blue optical photons. Therefore, on planets orbiting stars with strong FUV and weak NUV flux, a substantial O₃ atmosphere may arise via photochemical processes alone (Segura et al. 2010; Hu et al. 2012; Gao et al. 2015; Harman et al. 2015).

This strong photochemical source of O₃ may be detectable by future space observatories designed to carry out direct atmospheric spectroscopy of rocky planets (e.g., the HDST or LUVOIR mission concepts), and may be misinterpreted as evidence for biologic activity on these worlds. Therefore, characterization of the stellar FUV/NUV ratio is an essential complement to spectroscopy of exoplanet atmospheres to control for potential false-positive "biomarkers." Furthermore, it has been shown that the abundances of water and ozone, as well as the atmospheric equilibrium temperature, can respond to changes in the stellar flux on timescales ranging from minutes to years (Segura et al. 2010). Detailed knowledge of the absolute flux level and temporal behavior stellar spectrum is important for understanding the evolution of potentially habitable atmospheres.

X-ray and EUV Irradiance: Atmospheric Heating and Mass-loss—The ratios of X-ray to total luminosity of M dwarfs are orders of magnitude higher (≳10–100 ×) than those of the present day Sun (Poppenhaeger et al. 2010), and the smaller semimajor axes of the HZ around M dwarfs means that X-ray effects on HZ planets will likely be more important than on HZ planets orbiting solar-type stars (Cecchi-Pestellini et al. 2009). Soft X-ray heating of planetary atmospheres enhances evaporation and atmospheric escape (Scalo et al. 2007; Owen & Jackson 2012), which may impact the long-term stability of an exoplanetary atmosphere. Recent works suggest that the influences of early evolution of low-mass stars and XUV heating could lead to a bi-modal distribution of water fractions on Earth mass planets in the HZ of M dwarfs (Tian & Ida 2015). In order to model the atmosphere as a system, we require inputs for both heating (soft X-ray and EUV, see below) and photochemistry (FUV and NUV).

Extreme-UV (EUV; 100 ≲ λ ≲ 911 Å) photons from the central star are an important source of atmospheric heating and ionization on all types of extrasolar planets. For terrestrial atmospheres, increasing the EUV flux to levels estimated for the young Sun (≈1 Gyr; Ayres 1997) can increase the temperature of the thermosphere by a factor of ≳10 (Tian et al. 2008), potentially causing significant and rapid atmospheric mass-loss. Ionization by EUV photons and the subsequent loss of atmospheric ions to stellar wind pick-up can also drive extensive atmospheric mass-loss on geologic timescales (e.g., Rahmati et al. 2014 and references therein). Estimates of the incident EUV flux are therefore important for evaluating the long-term stability of an HZ atmosphere; however, a direct measurement of the EUV irradiance from an exoplanetary host star is essentially impossible because interstellar hydrogen removes almost all of the stellar EUV flux for virtually all stars except the Sun. The stellar EUV energy budget contains contributions from both the transition region (Lyman continuum as well as helium and metal line emission in the 228–911 Å bandpass) and the corona. FUV emission lines (Lyα, C iv, and Si iv) are required to estimate the contribution of the transition region to the EUV flux (Fontenla et al. 2011; Linsky et al. 2014), while X-ray data are necessary to constrain the contribution of the corona (e.g., Sanz-Forcada et al. 2011).

An important measurement relating to the habitability of extrasolar planets is the time variability of the energetic incident radiation. While most of the quiescent UV emission from M dwarfs comes from emission lines, continuum emission can become the dominant UV luminosity source during flares (Kowalski et al. 2010). The relative UV emission line strengths also vary during flares (e.g., Hawley et al. 2003; Osten et al. 2005; Loyd & France 2014). Thus, molecular species in the atmospheres of HZ planets will be "selectively pumped" during quiescent periods; only species that have spectral coincidences with stellar emission lines will be subject to large energy input from the host star. However, during flares with strong continuum emission, the relative excitation and dissociation rates relative to quiescent periods could change radically. Therefore, temporally and spectrally resolved observations are essential for understanding the impact of time variability on HZ planetary atmospheres. The amplitude and frequency of flare activity on older M-star exoplanetary hosts is completely unexplored, although GALEX NUV imaging observations suggest that flares may significantly alter the steady-state chemistry in the atmospheres of planets in the HZ (Welsh et al. 2006).

Impulsive UV events are also signposts for energetic flares associated with large ejections of energetic particles. Segura et al. (2010) have shown that energetic particle deposition into the atmosphere of an Earth-like planet without a magnetic field during a large M dwarf flare can lead to significant atmospheric O₃ depletion (>90% for the most extreme flares). Buccino et al. (2007) also studied the impact of a series of lower intensity flares from highly active stars. These events could alter the atmospheric chemistry and increase the penetration depth of UV photons that are damaging to surface life. The impact of a single flare may be detrimental to the development and maintenance of life, but the potentially far more significant impact of persistent flare events has not been studied because the temporal behavior of UV flares is unexplored outside of a few extreme M dwarf flare stars (e.g., AD Leo, EV Lac, and AU Mic; e.g., Hawley et al. 1991; Robinson et al. 2001; Osten et al. 2005).

With the previously described motivation in mind, the question that arises is "what are the shapes and absolute flux levels of the UV stellar SEDs incident on these planetary systems?" At present, GJ 832 is the only M dwarf for which a semi-empirical atmosphere model has been built and vetted by direct comparison with spectroscopic observations at optical, NUV, FUV, and X-ray wavelengths (Fontenla et al. 2016). There are no other stellar atmosphere models for M dwarfs that treat the chromosphere, transition region, and corona in a self-consistent manner, and none that can produce synthetic spectra for the important X-ray and ultraviolet bandpasses (5–3000 Å).

Several approaches have been taken in the literature, including assuming that the star has no UV emission (essentially, no magnetically active upper atmosphere; Segura et al. 2005; Kaltenegger et al. 2011), assuming the extreme flare environment of a star like AD Leo (Segura et al. 2010; Wordsworth et al. 2010), or using low signal-to-noise ratio (S/N) observations of the few flaring M dwarfs that could be observed by IUE (Segura et al. 2005; Buccino et al. 2007). Previously available archival data are insufficient for an accurate quantitative analysis of the radiation fields incident on potentially habitable planets orbiting M dwarfs. Low-sensitivity and contamination by geocoronal Lyα emission make IUE observations insufficient for this work (see example in Section 2). The lack of observational constraints from the UV spectra of M dwarfs will have a major impact on how we judge whether or not the planets in these systems are actually inhabited. While the need for panchromatic data has been realized for Sun-like stars (e.g., Sun in Time; Ribas et al. 2005), M dwarfs have received less attention (see also Guinan & Engle 2009), despite the fact that these systems dominate the planet statistics of the Galaxy.

To address the above question, we have carried out the first panchromatic survey of M and K dwarf exoplanet host stars in the solar neighborhood (d ≲ 20 pc). We refer to this program as the Measurements of the Ultraviolet Spectral Characteristics of Low-mass Exoplanetary Systems (MUSCLES) Treasury Survey, a coordinated X-ray to NIR observational effort to provide the exoplanet community with empirically derived panchromatic irradiance spectra for the study of all of the types of exoplanets orbiting these stars. MUSCLES is largely based on a Hubble Space Telescope (HST) Cycle 22 treasury program and makes use of smaller guest observing programs on HST, XMM-Newton, Chandra, and several ground-based observatories.

Our survey provides a database of the chromospheric, transition region, and coronal properties of low-mass stars hosting exoplanets, providing high-quality input for models of both Jovian and Earth-like planets as vast numbers of these systems are discovered and characterized in the next decade with missions such as WFIRST, Plato, TESS, and the James Webb Space Telescope (JWST). While the long-term evolution of the broadband UV luminosity function of M dwarfs can be constrained with large photometric samples from GALEX (e.g., Shkolnik & Barman 2014), a further uncertainty in the temporal behavior of low-mass exoplanet host stars is the variability on timescales of years (the stellar equivalent of the solar cycle). Given the limitations on space observatories, particularly as it is likely that we will be in the post-HST and post-Chandra era within the next 5–10 years, it is critically important to identify a set of visible-wavelength tracers (e.g., Gomes da Silva et al. 2011) that can be used to quantify the longer-term (years-to-decades) UV variability of these stars. The MUSCLES data set will enable us to derive empirical relations between the optical, UV, and X-ray fluxes (e.g., FUV luminosity versus Ca ii and Hα profiles), as well as their relative behavior during flares, supporting long-term, ground-based programs to study the time evolution of the energetic radiation environment.

This paper provides an overview of the motivation for and design of the MUSCLES Treasury Survey, as well as some initial quantitative results. A detailed example of the need for HST to carry out this work is given is Section 2. Section 3 describes the MUSCLES target list and the description of the observing modes used in the program. Section 4 presents the evolution of the broadband SED with stellar effective temperature and HZ distance, provides scaling relations to estimate the broadband luminosity from individual spectral line measurements, and presents a first look at the intense high-energy flare behavior of these otherwise inactive stars. Section 5 explores the interaction of the planets and host stars in these systems, and compares the UV flux measurements with predictions from coronal models. Section 6 presents a summary of the important results from this paper.

Details on all of the aspects of the data analysis and scientific results from the program can be found in the companion papers from the MUSCLES team. Youngblood et al. (2016) present an analysis of the essential reconstruction of the intrinsic Lyα stellar emission lines and the calculations of the EUV irradiance ("Paper II"). Loyd et al. (2016), "Paper III," describe the creation of the panchromatic SEDs, quantification of the SEDs for exoplanet atmospheric modeling, and a description of how to download the data in machine-readable format from the Milkulski Archive for Space Telescopes.²⁰ Fontenla et al. (2016) present semi-empirical modeling of M dwarf atmospheres based in part on MUSCLES observations. Detailed descriptions of the flare properties of the MUSCLES sources and a comparison of the UV and X-ray emission from these stars with contemporaneous ground-based photometry and spectroscopy are presented in papers by R. O. P. Loyd (2016, in preparation) and A. Youngblood (2016, in preparation), respectively, and J. Linsky et al. (2016, in preparation) describe the kinematics of the stellar atmospheres derived from the UV emission lines in the HST data.

The MUSCLES target list (Table 1) was chosen to cover a broad range of stellar types (K1V–M5V; 4 K dwarfs and 7 M dwarfs), exoplanet masses, and semimajor axes, including most of the known M dwarf exoplanet host stars located within d ≲ 13 pc (7/12), while excluding flare stars (e.g., GJ 674) that require intensive multi-wavelength monitoring to clear HST instrument safety protocols. The M dwarfs span a range of spectral types (from M1 to M5), a range of X-ray luminosity fractions (log₁₀(L_X/L_Bol) ≈ −5.1 to −4.4, an indicator of activity level), and planetary systems ranging from Jupiters (GJ 832) to super-Neptunes (GJ 436) to super-Earths (GJ 1214). About ∼65% of the exoplanet host stars in our sample (7/11) harbor super-Earths (M_plan < 10 M_⊕; bold in Table 1). In the brief summaries of the star–planetary systems given in the Appendix, we refer to the M sin iof the planets as their mass as a shorthand.

Table 1.  MUSCLES Treasury Survey—Target List

Star	Distance	Type	Teff	Prot	$\langle {r}_{\mathrm{HZ}}\rangle$ a	Exoplanet Mass	Semimajor Axis	Ref.
	(pc)		(K)	(days)	(au)	M sin i (M⊕)	(au)
GJ 1214	14.6	M4.5	2935	53	0.096	6.4	0.0143	1, 2, 3
GJ 876	4.7	M5	3062	96.7	0.178	615, 194,	0.208, 0.130,	3, 4, 5
						5.7, 12.4	0.021, 0.334
GJ 581	6.3	M5	3295	94.2	0.146	15.9, 5.4,	0.041, 0.073,	3, 4, 6
						6.0, 1.9	0.218, 0.029
GJ 436	10.3	M3.5	3281	48	0.211	23	0.0287	3, 4, 7
GJ 176	9.4	M2.5	3416	38.9	0.262	8.3	0.066	3, 4, 8
GJ 667C	6.9	M1.5	3327	105	0.172	5.7, 4.4	0.049,0.123	3, 1, 9
GJ 832	4.9	M1.5	3816	40b	0.235	203, 5.4	3.6, 0.16	3, 4
HD 85512	11.2	K6	4305	47.1	0.524	3.5	0.26	3, 10, 11
HD 40307	12.9	K2.5	4783	48	0.674	4.1, 6.7,	0.047, 0.080,	3, 10, 12
						9.5, 3.5,	0.132, 0.189,
						5.1, 7.0	0.247, 0.600
Eri	3.2	K2	5162	11.7	0.726	∼400	3.4	3, 10, 13
HD 97658	21.1	K1	5156	38.5	0.703	6.4	0.080	3, 14, 15

Notes.

^aThe habitable zone distance is defined as the average of the "runaway greenhouse" and "maximum greenhouse" limits (Kopparapu et al. 2014). ^bGJ 832 has no published rotation period, we assume a relatively short period due to the persistence of UV flare activity in this star.

References: 1. Neves et al. (2014), 2. Berta et al. (2011), 3. Simbad parallax distance, 4. Gaidos et al. (2014), 5. Rivera et al. (2010), 6. Robertson et al. (2014), 7. Demory et al. (2007), 8. Kiraga & Stepien (2007), 9. Anglada-Escudé et al. (2013), 10. Tsantaki et al. (2013), 11. Pepe et al. (2011), 12. Mayor et al. (2009), 13. Donahue et al. (1996), 14. Valenti & Fischer (2005), 15. Henry et al. (2011).

Download table as:  ASCIITypeset image

With the exception of Eri, the MUSCLES stars are not traditionally classified as active or flare stars (in contrast with widely studied M dwarf flare stars such as AU Mic or AD Leo). The stars in our sample are considered to be "optically inactive," based on their Hα absorption spectra (Gizis et al. 2002). However, the Ca ii H and K emission line cores are a more straightforward means of diagnosing chromospheric activity in the optical (Gomes da Silva et al. 2011) because the Ca ii emission line flux increases with activity while Hα first shows enhanced absorption before becoming a strong emission line with increasing activity. All of our stars with measured Ca ii H and K profiles show weak but detectable emission (equivalent widths, W_λ(Ca ii) > 0; note that the references below present emission lines as positive equivalent widths, which is in opposition from the traditional convention), indicating that at least a low level of chromospheric activity is present in these stars (Rauscher & Marcy 2006; Walkowicz & Hawley 2009). In addition, Hawley et al. (2014) have shown that inactive M dwarfs display flares in their Kepler light curves, confirming that chromospheric activity is present on MUSCLES-type stars. Figure 3 shows the Ca ii K line equivalent width as a function of $(B-V)$ color for six of the M dwarfs in the sample, overplotted on the data from Rauscher & Marcy (2006). The Ca ii K emission is in the range 0.1 < W_λ(Ca ii) < 0.8 Å, which is approximately an order of magnitude smaller than traditional flares stars EV Lac, AU Mic, AD Leo, and Proxima Cen (also shown in Figure 3). According to the M dwarf classification scheme of Walkowicz & Hawley (2009), these stars have intermediate chromospheres and are referred to as "weakly active." In the Appendix, we present brief descriptions of each of the stars studied here. For more detailed descriptions of the stellar parameters of the MUSCLES target stars, we refer the reader to Loyd et al. (2016) and Youngblood et al. (2016).

Zoom In Zoom Out Reset image size

Ultraviolet Observations, 100–3200 Å—In order to obtain a full census of the UV emission incident on the HZs of low-mass stars, we require HST spectral coverage from 1150–3100 Å: the G130M, G160M, and G230L modes of COS, and the G140M, E140M, E230M, E230H, and G230L modes of STIS provide spectral coverage across this bandpass. This combination of instrumental settings allows us to catalog the stellar emission lines that are relevant to the photoexcitation of the primary atmospheric constituents of the exoplanets in these systems. The H i Lyα emission line dominates the total UV luminosity of M dwarfs (France et al. 2012a). We use the G140M mode of STIS with the 52'' × 01 slit to measure the Lyα profile for our M dwarf targets. We have previously demonstrated that this technique can produce high-quality measurements of the local Lyα flux (France et al. 2013). For the brighter K dwarfs, we employ the STIS E140M mode, using the 02 × 006 slit to resolve the intrinsic line profile and minimize contamination by telluric H i. Resonant scattering of Lyα in the local ISM requires that the line must be reconstructed to provide a reliable measure of the intrinsic Lyα radiation field in these exoplanetary systems. We direct the reader to Paper II for a detailed description of the Lyα reconstruction developed for the MUSCLES Treasury Survey (Youngblood et al. 2016).

In the FUV (except Lyα), we use COS G130M (Si iiiλ1206, N vλ1240, C iiλ1335, and Si ivλ1400 lines) and COS G160M (C ivλ1550, He iiλ1640, and Al iiλ1671 lines). The emission from Lyα, Si iv, C iv, and He ii is particularly interesting because these lines provide constraints on the Lyman continuum/EUV (200 ≲ λ ≲ 900 Å) irradiance in these systems (Linsky et al. 2014; Shkolnik et al. 2014). COS is essential for a moderate spectral resolution FUV line census as the lower effective area and higher detector background of STIS make observations of all but the very brightest emission line (H i Lyα) prohibitively time consuming. We use the medium-resolution ("M") COS modes to resolve chromospheric emission lines and maximize contrast from narrow spectral features.

At NUV wavelengths, we use STIS G230L (λ > 2200 Å) to observe the NUV continuum, Fe iiλ2400 and λ2600, and Mg iiλ2800, but take advantage of the superior sensitivity of the COS G230L mode to observe the 1750–2200 Å region that is important for the photodissociation of O₂ and the production of O₃. For targets that exceed the G230L bright-object limit (the K stars, Eri, HD 40307, HD 85512, and HD 97658), we use the higher-resolution E230M mode (λ1978 + λ2707 settings) covering the 1800–3100 Å bandpass. Target brightness limits dictated that we employ the STIS E230H mode for observations of the Mg ii emission lines on the brightest K dwarf ( Eri).

The Lyα fluxes are combined with models of solar active regions (Fontenla et al. 2011) to estimate the EUV luminosity in the wavelength region 100–1170 Å in 100 Å bins (see Linsky et al. 2014 and Paper II). These EUV calculations were compared to the GJ 832 atmosphere model developed by Fontenla et al. (2016); we found that the 10–100 Å and 100–1000 Å bands differed by less than a factor of two. Given that no inactive M dwarf EUV observations exist as a reference point, we consider this intermodel agreement to be sufficient.

The MUSCLES observing plan was designed to study both the spectral and temporal variability of low-mass exoplanet host stars on the characteristic timescale for UV/optical variability (Kowalski et al. 2009). Using the most sensitive UV photon-counting mode on HST (COS G130M) and the time-tag capability of the COS microchannel plate detector (Green et al. 2012), we measure the 10^4–5 K chromospheric and transition region activity indicators (using the C ii, Si iii, Si iv, and N v emission lines) in 8 hr intervals (5 contiguous spacecraft orbits for each star). Quasi-simultaneous X-ray observations were coordinated with these visits when possible (see below). While this strategy does not provide continuous coverage due to Earth occultation, it is optimized for constraining the types and frequency of flare behavior on low-mass exoplanet host stars. Since the characteristic timescale for UV/optical flare activity on M dwarfs is thought to be minutes to hours (Welsh et al. 2007; Kowalski et al. 2009; Loyd & France 2014), our observations are ideal for quantifying the importance of flare activity to the local UV radiation field in these systems.

With the above considerations in mind, we arranged the UV—visible observations for each target into campaigns comprising three HST visits: (1) COS G130M, (2) COS G160M+G230L, and (3) STIS G/E140M+G230L+G430L, all executed within a day of each other to mitigate uncertainties introduced by month or year timescale variations in the stellar flux. Between 9 and 13 total HST exposures were acquired for each star depending on the target brightness and the observing modes used. For a graphical description of which modes contribute at which wavelengths for each target, we refer the reader to Loyd et al. (2016).

X-ray Observations, 5–50 Å (2.5–0.25 keV)—We used quasi-simultaneous Chandra and XMM-Newton observations of the MUSCLES stars to provide temporally consistent SEDs and to explore the wavelength-dependent behavior of stellar flares. The X-ray observations were coordinated with the COS G130M 5-orbit monitoring program (see above) to establish the longest possible simultaneous baseline over which to explore the panchromatic properties of M and K dwarf flares. We observed GJ 667C, GJ 436, GJ 176, and GJ 876 using the Chandra ACIS-S back-illuminated S3 chip. Owing to an HST safing event in 2015 June, the GJ 876 campaign was not simultaneous. We observed Eri and GJ 832 with XMM-Newton because they are optically bright. Eri is by far the brightest X-ray source in our sample and provided a high-quality RGS grating spectrum. Eri required use of the pn/MOS Thick filters, while GJ 832 used the Medium filters. We also obtained photometry with the UVM2 filter for GJ 832. HD 85512 and HD 40307 were observed as part of a complementary XMM-Newton program (PI—A. Brown) using the same configuration as for GJ 832. For the stars without contemporaneous X-ray data, we obtained spectra from the Chandra and XMM-Newton archives (A. Brown et al. 2016, in preparation; Loyd et al. 2016). The coronal model fit to the X-ray data is extended into the 50–100 Å region where no direct observations are available.

Optical Observations, 3200–6000 Å—We carried out complementary ground-based observations as close in time as possible to the MUSCLES UV/X-ray observations. Our primary optical spectra came from the ARCES and DIS instruments on the Astrophysical Research Consortium (ARC) 3.5 m telescope at Apache Point Observatory (3700–10000 Å, depending on the mode), the 2.15 m Jorge Sahade telescope at the Complejo Astronomico El Leoncito (Cincunegui & Mauas 2004), and the FLOYDS intrument on the Las Cumbras Observatory Global Telescope Network (Brown et al. 2013). Multi-band optical photometry was acquired with the Apache Point Observatory 0.5 m ARCSAT telescope (with the Flarecam instrument) and the LCOGT. The optical data and their correlation with the spectral and temporal behavior of the high-energy emission will be presented in a future work by the MUSCLES team (A. Youngblood et al. 2016, in preparation). In order to calibrate the UV data with respect to visible/IR photospheric models and ground-based spectra, we also acquired short optical observations with STIS G430L or G430M (depending on target brightness) during each HST campaign.

Infrared Extension, >6000 Å—Stellar atmosphere models considering only the photosphere (e.g., the PHOENIX models) are able to reproduce the peak and Rayleigh–Jeans tail of the stellar SED given the correct prescription for the effective temperature and abundances of atoms and molecules (Allard & Hauschildt 1995; Husser et al. 2013). We employ PHOENIX models, matched to the Tycho B- and V-band fluxes, to calibrate the optical spectroscopy from HST-STIS and fill out the panchromatic SEDs from the red-optical to the mid-IR (∼0.6–5.5 μm; Loyd et al. 2016). The red-optical and infrared fluxes are important as they make up the majority of the bolometric luminosity from cool stars, regulate the effective surface temperature of orbiting planets, and provide the reference for the activity indicators (e.g., L(N v)/L_Bol) presented in this work.

Figure 6 shows a montage of the full 5 Å–5 μm SEDs for the 11 MUSCLES exoplanet host stars, binned to 5 Å pixel⁻¹ for display and divided by the bolometric luminosity. The bolometric luminosity used in this work is simply the integral of the complete SED for each star (including a calculated Rayleigh–Jeans tail at λ > 5.5 μm; Loyd et al. 2016). Many of the component spectra have spectral resolutions much higher than this (resolving powers of R > 15,000 for all of the FUV data and up to R = 114,000 for STIS E230H observations of Eri), and both full-resolution and binned data sets are available on the MAST HLSP archive.²² As discussed in the Introduction, the broadband behavior of the stellar SED can provide general insights into the relevant heating and photochemical rates for all types of exoplanet atmospheres.

Zoom In Zoom Out Reset image size

We define the FUV flux, F(FUV), as the total stellar flux integrated over the 912–1700 Å bandpass, including the reconstructed Lyα emission line, FUV = (F(Lyα) + F(1170–1210 + 1220–1700) + F(912–1170)). F(Lyα) is the reconstructed Lyα line flux, F(1170–1210 + 1220–1700) is the non-Lyα flux directly measured by HST-COS and STIS, and F(912–1170) is the "Lyman Ultraviolet" emission constructed from the solar active region relations for low-mass stars presented by Linsky et al. (2014). On average, Lyα contributes 83% (±5%) to the total FUV flux. F(NUV) are the combined HST spectra integrated over 1700–3200 Å. The fractional luminosities in each band as well as key emission lines are given in Table 3.

Figure 7 shows the FUV/NUV flux ratio as a function of the stellar effective temperature. The FUV/NUV flux ratio is in the range 0.5–0.7 for the latest M stars (GJ 1214 and GJ 876), is in the range 0.2–0.4 for M1 to M3 stars, declines to 0.04 by mid-K, and is ≲0.01 for spectral types K2 and earlier. France et al. (2013) have previously shown that the FUV/NUV ratio drops to ∼10⁻³ for solar-type stars. The trend in FUV/NUV with effective temperature is largely driven by the large increase in photospheric emission moving into the NUV spectral bandpass with increasing effective temperature. Large FUV fluxes dissociate O₂ to generate O₃. The photospheric NUV photons destroy photochemically produced ozone and keep the atmospheric O₃ mixing ratios low on Earth-like planets around solar-type stars in the absence of a disequilibrium process such as life.

Zoom In Zoom Out Reset image size

The XUV flux (5–911 Å), the combination of the observed and modeled soft X-ray flux (usually 2.5–0.125 keV, or ≈5–100 Å)²³ and the calculated EUV flux (100–911 Å), is an important heating agent on all types of planets. XUV irradiance is particularly important for short-period planets (Lammer et al. 2009). Figure 8 shows the relative XUV and FUV luminosities as a function of effective temperature for the MUSCLES stars. The total XUV and FUV luminosities are shown to be well-correlated with stellar effective temperature (and therefore stellar mass) and the fractional XUV and FUV luminosities are in the range 10⁻⁵ ≲ L(band)/L_bol ≲ 10⁻⁴ with no dependence on stellar effective temperature. For comparison, the disk-integrated quiet Sun (Woods et al. 2009) has L(XUV)/L_Bol = 2 × 10⁻⁶ and L(FUV)/L_Bol = 1 × 10⁻⁵, respectively. Note, however, that the solar L(FUV) contains a contribution from the photosphere, whereas the photospheric contribution is negligible for M and K dwarfs. From the perspective of integrated planetary atmosphere mass-loss over time, our measurements represent only a conservative lower-limit as the stellar XUV radiation was likely factors of 10–100 times higher during the star's younger, more active periods. The active periods of M stars are more prolonged than the equivalent "youthful magnetic exuberance" (Ayres & France 2010) period of solar-type stars (e.g., West et al. 2008). Additionally, it is likely that the stellar mass-loss rates are higher earlier in their evolution (Wood et al. 2005a), increasing the potential planetary mass-loss.

Zoom In Zoom Out Reset image size

We can combine these results into a comprehensive picture of the energetic radiation environment in the HZs around the low-mass stars. Figure 9 shows the FUV/NUV ratios, the total FUV fluxes, and the total XUV fluxes at the HZ orbital distances for each of our targets. The plots show the average HZ distance for each star, $\langle {r}_{\mathrm{HZ}}\rangle$, computed as the mean of the "runaway greenhouse" and "maximum greenhouse" limits to the HZ presented by Kopparapu et al. (2014). The error bars on $\langle {r}_{\mathrm{HZ}}\rangle$ represent these extrema. One observes the two orders of magnitude decline of the FUV/NUV ratio from 0.1 to 0.7 au, mainly driven by the stellar effective temperature dependence described above. The FUV HZ fluxes show a weak trend of increasing flux with increasing r_HZ, however, both the M dwarf and K dwarf samples have a factor of roughly 5 dispersion at a given HZ distance. For example, while it is possible for the absolute FUV flux to be a factor of 10 greater at 0.7 au than at 0.15 au, it is also possible that the FUV flux at 0.15 au is greater than at 0.7 au. A similar dispersion is seen in the XUV fluxes, and there is no statistically significant change in the XUV flux across the HZ (Table 2). Therefore, the average FUV and XUV fluxes in the HZs of M and K dwarfs are 10–70 erg cm⁻² s⁻¹. We emphasize that the activity level of the individual star must be considered, and direct observations are preferable when available. In Section 4.3, we will discuss the correlation of the broadband fluxes with specific emission line measurements to simplify the characterization of the broadband fluxes from single emission line flux measurements.

Zoom In Zoom Out Reset image size

Table 2.  Emperical Log–Log Relations, log₁₀ y = (m × log₁₀x) + b

y	x	m	b
L(FUV)a	L(C iv)	1.12 ± 0.27	−1.01 ± 6.86
L(FUV)	L(Mg ii)	0.75 ± 0.06	7.56 ± 1.52
L(XUV)	L(C iv)	0.97 ± 0.19	2.72 ± 4.91
L(XUV)	L(Mg ii)	0.58 ± 0.08	11.98 ± 2.12
FUV/NUV	$\langle {r}_{\mathrm{HZ}}\rangle$	−2.42 ± 0.30	−2.31 ± 0.18
HZ FUV Fluxb	$\langle {r}_{\mathrm{HZ}}\rangle$	0.51 ± 0.16	1.71 ± 0.10
HZ XUV Fluxb	$\langle {r}_{\mathrm{HZ}}\rangle$	−0.01 ± 0.22	1.41 ± 0.14
L(N v)/LBol	Mplan/aplan	0.64 ± 0.24	−8.53 ± 0.56

Notes.

^aAll luminosities are in units of (erg s⁻¹). ^bHabitable zone fluxes are in units of (erg cm⁻² s⁻¹), $\langle {r}_{\mathrm{HZ}}\rangle$ and a_plan in units of (au), M_plan in units of (M_⊕).

Download table as:  ASCIITypeset image

Chromospheric and transition region emission lines are observed in all of the MUSCLES spectra, which argues that all exoplanet host stars with spectral type M6 and earlier have UV-active atmospheres (Figures 4 and 5). This seems to rule out photosphere-only models of cool stars and indicates that the chromospheric, transition region, and coronal emission must be included for accurate modeling of the atmospheres of planets orbiting these stars. Bright emission lines in the MUSCLES spectra with chromospheric formation temperatures in the range ∼(4–30) × 10³ K include H i Lyα, Si iiλ1264, 1526, 1808, C iiλλ1334, 1335, Al iiλ1671, Fe ii multiplets near 2400 and 2600 Å, and Mg ii λλ2796, 2803 (see, e.g., the M dwarf contribution functions presented by Fontenla et al. 2016). We observe many transition region lines with formation temperatures in the range ∼(40–200) × 10³ K, including the C iii 1175 multiplet, Si iiiλ1206, O vλ1218, 1371, N vλλ1239, 1243, O iv] λ1401, Si ivλλ1394, 1403, C ivλλ1548, 1550, and He iiλ1640. The coronal iron lines, Fe xiiλ1242 and Fe xxiλ1354 (Ayres et al. 2003), are observed in a fraction of HST-COS observations (7/11 stars show Fe xii emission lines while only GJ 832 and GJ 876 show Fe xxi emission lines). Taken together with the X-ray observations, these highly ionized iron lines demonstrate the presence of coronal gas in the atmospheres of all of our stars and enable an alternative calculation of the EUV irradiance using emission measure techniques (Sanz-Forcada et al. 2011; Chadney et al. 2015). We will present a detailed discussion of the atmospheric kinematics derived from emission line parameters in an upcoming work (J. Linsky et al. 2016, in preparation).

A major complication in previous attempts to assemble panchromatic radiation fields, particularly those of M dwarfs, is the flux variability between observations separated by years (e.g., comparing IUE spectra from the 1980s with ROSAT X-ray data from the 1990s and HST observations of Lyα from the 2000s; Linsky et al. 2013, 2014). One of the goals of the MUSCLES Treasury Survey is to obtain multi-wavelength observations close in time to minimize this large systematic uncertainty. In the MUSCLES observing strategy, emission lines with similar formation temperatures were acquired on different HST visits (due to their inclusion on different HST grating modes) separated by 18–48 hr. The overlap in formation temperature and spectral coverage between adjacent modes facilitates scaling over calibration variations (Loyd et al. 2016) and smaller day-to-day variations.

This approach has been successful—Figure 10 shows the relationship of the fractional luminosity in the transition region ions N v and C iv. These lines have formation temperatures within a factor of two of each other (between (1–2) × 10⁵ K) and have been shown to be tightly correlated in numerous astrophysical plasmas, including the atmospheres and accretion columns around young stars (Oranje 1986; Ardila et al. 2013; France et al. 2014). The N v versus C iv correlation is very well maintained over the MUSCLES sample despite the non-simultaneous observations, with a Pearson correlation coefficient of 0.90 (ρ in the legend of Figure 10) and a probability of no correlation of 6.0 × 10⁻⁵ (n in the legend of Figure 10). Because we do not exclude discrete impulsive flares, the MUSCLES spectra can be considered to be an accurate snapshot of the average stellar spectrum (averaged on the timescale of hours-to-days). Variability on the timescales of the solar cycle (or stellar cycles, years) will be addressed by developing XUV–optical tracer correlations and carrying out long-term monitoring from ground-based facilities (Section 1.3).

Zoom In Zoom Out Reset image size

The emission lines in the MUSCLES stars also show an evolution of decreasing fractional luminosity of the transition region with increasing effective temperature. Figure 11 (top) shows the fractional N v luminosity as a function of the effective temperature. Given that the total XUV and FUV factional luminosities are approximately constant with effective temperature (Figure 8, lower), this suggests that the upper transition region activity (traced by N v) declines faster than the rest of the upper atmosphere (e.g., the chromosphere and corona), which may be related to the pressure-density structure of the stellar atmosphere changing with mass (Fontenla et al. 2016). While the Pearson correlation coefficient is only −0.60 (n = 3.5 × 10⁻²), this is skewed by the inclusion of Eri, the only active star in the sample. Excluding Eri, the coefficient becomes −0.65 (n = 9.1 × 10⁻³). The middle plot in Figure 11 shows the total N v luminosity with the stellar rotation rate, suggesting a period–activity relation analogous to the well-studied relationship in solar-type stars (Ribas et al. 2005; see also Engle & Guinan 2011 for M dwarf X-ray evolution). The notable exception is GJ 876, with an N v luminosity approximately an order of magnitude larger than expected based on its rotational period. This is partially the result of the strong flare activity on this star (see Section 4.4), and also suggests that the period is not well determined for this star. Excluding GJ 876, the Pearson and Spearman coefficients are −0.62 and −0.92, respectively. If GJ 876 fell on this trend, then we would expect a rotation period closer to ∼40 days. Interestingly, there is no trend in the fractional N v luminosity with stellar rotation rate (Figure 11, bottom).

Zoom In Zoom Out Reset image size

Another goal of the MUSCLES Treasury Survey is to identify individual spectral tracers that can serve as proxies for the broadband fluxes from low-mass stars so that large, resource-intensive projects may not be required to obtain accurate estimates for the energetic radiation environments around exoplanets in the future. This anticipates a time beyond the current suite of UV and X-ray observatories (e.g., XMM, Chandra, and HST) capable of making these measurements, and a time when the sheer number of potentially habitable planets around low-mass stars precludes a detailed panchromatic characterization of every target (possibly after the TESS mission). In this case, simple relationships for acquiring reasonably accurate broadband stellar flux measurements will be essential for accurately modeling the atmospheric spectra from these worlds when they are acquired by future flagship missions in the 2020s, 2030s, and 2040s. Below, we describe the relationship between the XUV and FUV luminosities of the MUSCLES stars with the most prominent FUV and NUV emission lines, Mg ii and C iv.²⁴ In a future work, we will explore ground-based tracers of the UV radiation environment (e.g., Ca ii H & K fluxes and equivalent widths from our contemporaneous ground-based observations; A. Youngblood et al. 2016, in preparation).

Figure 12 shows the relationship between the broadband luminosities L(XUV) and L(FUV) and the emission line luminosities L(C iv) and L(Mg ii). The figures show a strong correlation between all of these quantities, with Pearson correlation coefficients of [0.83, 0.86, 0.87, 0.90] for the relationships [FUV–C iv, FUV–Mg ii, XUV–C iv, XUV–Mg ii)], respectively. The probability of a non-correlation is <3 × 10⁻³ for all four curves. We present quantitative log-log relations²⁵ in Table 2. The rms scatter around the [FUV–C iv, FUV–Mg ii, XUV–C iv, XUV–Mg ii)] fit, (L(band) –L(fit))/L(band), is [153%, 39%, 109%, 57%], respectively.

Zoom In Zoom Out Reset image size

We have selected C iv and Mg ii because they are the two most readily-observable emission lines in the FUV and NUV bandpasses. Mg ii shows a tighter correlation, but as we begin to probe exoplanetary systems at greater distances from the Sun, this relationship will become compromised by the additional contribution from interstellar Mg ii absorption components with radial velocities coincident with the stellar radial velocities (Redfield & Linsky 2002) as well as the possibility of gas-rich circumstellar environments fueled by mass-loss from short-period gaseous planets (Haswell et al. 2012; Fossati et al. 2015). By contrast, C iv is essentially free from interstellar extinction out to the edge of the Local Bubble where dust opacity begins to contribute appreciably. There are rare exceptions where hot gas in the Local Bubble can contribute small amounts of C iv attenuation (Welsh et al. 2010), but the effect is considerably less than for Mg ii.

As discussed in the Introduction, temporal variability of the energetic radiation is considerably higher for M dwarfs than for solar-type stars (Mitra-Kraev et al. 2005), with active M stars showing disk-integrated flux increases of an order of magnitude or more during large flares (Hawley & Pettersen 1991; Hawley et al. 2003; Osten et al. 2005). The MUSCLES pilot program showed that even inactive M dwarfs could show impulsive flare behavior in their UV light curves (France et al. 2012a). The UV monitoring component of the MUSCLES Treasury Survey was designed to present a uniform database of flares from exoplanet host stars so that the flare frequency-amplitude relations could be derived for these stars and the impact of impulsive events on the atmospheres or orbiting planets could be assessed. For cases where the X-ray observations were scheduled simultaneously, the overlapping observatory coverage was planned for the HST 5-orbit monitoring campaigns.

The intention of this section is not to give a thorough quantitative description of the flare catalog produced in the survey; that work will be presented in a follow-up paper by R. O. P. Loyd et al. (2016, in preparation). This section is intended to introduce the variability data and present one example of an optically quiet M dwarf host star that is among the most UV/X-ray active sources ever observed. Figure 13 shows the light curves of GJ 876, recorded in four bright emission lines, C ii λλ1334, 1335 (T_form ∼ 3 × 10⁴ K), Si iiiλ1206 (T_form ∼4 × 10⁴ K), Si ivλλ1394, 1403 (T_form ∼ 6 × 10⁴ K), and N vλλ1239, 1243 (T_form ∼ 2 × 10⁵ K). The data are binned to a 30 s cadence and displayed relative to the start time of the first observation. The individual exposures are labeled for reference. Several flares are immediately apparent, the strongest being during the last two orbits of the monitoring campaign. The COS detector background, measured at the same dispersion direction location as the Si iv lines but offset below the spectral trace in the cross-dispersion direction, is also shown (orange squares) to demonstrate the stability of the instrument during these measurements.

Zoom In Zoom Out Reset image size

Figure 14 (top) shows a zoom-in on the brightest UV flare, binned to a 10 s cadence, occurring near T_exp = 17,400 s. Each light curve is normalized to unity during a pre-flare window (17,100–17,250 s) to enable a comparison of the relative flare responses of each line. The flare/quiescent flux increase in the brightest line, Si iii, is ∼110. C ii and Si iv show flux increases of the order of ∼50, and N v and the FUV continuum increase by factors of ∼5. This level of luminosity increase indicates the largest relative UV flare ever directly detected in a disk-integrated observation of a star, despite the relative inactivity of GJ 876 suggested by the Ca ii activity index. One observes that the light curves are line-dependent, with intermediate temperature ions showing a larger relative flux increase while the higher temperature ion (N v) does not respond as strongly but shows a decay time several times longer than C ii, Si iii, and Si iv (Figure 13). Figure 14, (bottom) shows time-resolved line ratios in the GJ 876 light curves. Several emission lines are ratioed to Si iv and normalized to their pre-flare line ratios, and so their relative change is meaningful. In this manner, we can place constraints on the atmospheric temperature regime where most of the observable energy is deposited (our observations do not contain information about energy deposited in cooler or optically thick atmospheric layers). The Si iii/Si iv ratio increases during the flare, while the C ii/Si iv and N v/Si iv ratios, which sample gas both hotter and colder than Si iii, are depressed. This argues that the flare energy distribution is peaked near the Si iii formation temperature, roughly (4–5) × 10⁴ K (with both low and high-energy tails), and evolves during the flare.

Zoom In Zoom Out Reset image size

In order to understand the impact of these flares on the planets in these systems, we need to convert the observed emission line flares to an estimate of the broadband UV flare energy. Converting the raw spectral light curves into flux units by comparing the orbit-averaged count rates and stellar emission line fluxes, we can create flux-calibrated light curves (Loyd & France 2014). The total flare energy in a given emission line is then

$$\begin{eqnarray}&&E(\lambda )=4\pi {d}^{2}{\displaystyle \int }_{{t}_{\mathrm{start}}}^{{t}_{\mathrm{end}}}F(\lambda ){dt},\end{eqnarray} \tag{ 1 }$$

where F(λ) is the line flux in (erg cm⁻² s⁻¹), d is the stellar distance, and t_start and t_end are the initial flare rise times and the time when the flare returns to the quiescent level, respectively. For the bright flare considered here, the exposure ends before the ionic lines return to the quiescent levels, and so we set t_end as the end of the orbit in this case. The Si iii and Si iv emission lines are representative of the upper chromosphere and transition region emission formed between ∼30 and 80 kK; we use these lines to estimate the total flare emission from these regions. The total energies in these lines are log₁₀ E(Si iii) = 29.17 erg and log₁₀ E(Si iv) = 29.08 erg. We use the M dwarf stellar atmosphere model of Fontenla et al. (2016) to estimate the total emission in the 300–1700 Å range that contains the majority of the transition region emission. Emission at λ < 300 Å is mainly coronal in origin while emission at λ > 1700 Å is mainly cooler chromospheric gas.

The ionized silicon flux energies can be converted to broadband XUV + FUV (300–1700 Å) energy by computing the fractional flux emitted in these lines and the fractional radiative cooling rate that is contributed by the 30–80 kK gas. The total broadband energy is then

$$\begin{eqnarray}&&E(300-1700\;\mathring{\rm A} )=E(\mathrm{Si}\;{\rm{III}}+\mathrm{Si}\;{\rm{IV}})\\ &&\quad \times \;({f}_{300-900}+{f}_{900-1200}+{f}_{1200-1700})\times {C}_{30-80\mathrm{kK},}\end{eqnarray} \tag{ 2 }$$

where f_300–900 is the ratio of stellar flux in the 300–900 Å band to the combined Si iii + Si iv flux, F(300–900 Å)/F(Si iii + Si iv). Similarly, f_900–1200 = F(900–1200 Å)/F(Si iii + Si iv) and f_1200–1700 = F(1200–1700 Å)/F(Si iii + Si iv). The f_300–900, f_900–1200, and f_1200–1700 ratios from the model atmosphere of Fontenla et al. (2016) are 32, 14, and 290, respectively. C_{30–80 kK} is the fraction of the total radiative cooling rate from the upper stellar atmosphere (6000–10⁶ K) contributed by 30–80 kK gas, C_{30–80 kK} = Γ(30–80 kK)/Γ(6000 K–10⁶ K), where Γ is the radiative cooling rate in units of erg cm⁻³ s⁻¹. Using the cooling rates for this model atmosphere (Fontenla et al. 2016), we compute C_{30–80 kK} = 16.8%. Note that these are the equilibrium cooling rates for the quiescent model atmosphere, and may be different during the post-reconnection heating associated with the flare. Combining these elements, we estimate that the total UV flare energy associated with this event is log₁₀ E(300–1700 Å) = 31.18 erg, which is comparable to the total quiescent luminosity of the star E_flare(UV) ∼ 0.3 L_*Δt (Δt = 1 s).

In a future work, R. O. P. Loyd et al. (2016, in preparation) will show that HST monitoring observations are able to measure and quantify the M dwarf flare amplitude–frequency distribution. Critically, HST provides the sensitivity and spectral resolution to analyze light curves covering a range of formation temperatures, and capture the short duration events that dominate the flare distribution. A dedicated HST spectroscopic flare monitoring program is currently the best avenue for understanding the energy and temporal distribution of flares and their potential influence on low-mass planets.

4.4.1. X-Ray Light Curves

When possible, X-ray observations were taken in concert with the UV light curves described here. However, this was only possible for about half of the MUSCLES observations. Figure 15 shows the non-simultaneous X-ray light curve of GJ 876,²⁶ finding another large flare on this object. The X-ray spectra of both the quiescent and flare periods have been fitted, and the overall X-ray luminosity increased by a factor of ∼10 during the flare. The actual luminosity increase was likely larger but diluted by the cadence adopted to provide sufficient S/N in each time bin. Given the similarity of the UV and X-ray light curves of dMe stars observed by XMM (Mitra-Kraev et al. 2005), it seems likely that these large flares are similar events and occur with regularity on GJ 876. We note that Poppenhaeger et al. (2010) also noted a high level of X-ray activity on GJ 876 and France et al. (2012a) observed a large UV flare in the very first MUSCLES observations. This apparently inactive star displays a flare outburst almost every time it is observed at wavelengths below the atmospheric cut-off. While we have elected to focus on the most UV-active star in the MUSCLES data set, roughly half of our targets displayed UV flare activity. This will be explored in greater detail in R. O. P. Loyd et al. (2016, in preparation).

Zoom In Zoom Out Reset image size

We use GJ 1214, the only star in our sample with an effective temperature of <3000 K, to investigate the evolution of upper atmosphere activity toward the stellar/sub-stellar boundary. In the MUSCLES pilot study, we presented an upper limit to the Lyα flux of GJ 1214 based on earlier STIS G140M observations. GJ 1214 has a spectral type of M4.5 V with T_eff = 2949 ± 30 K (Kundurthy et al. 2011; consistent with the 2935 K derived from the photospheric model fitting described in Paper III). It was the most distant source in the MUSCLES pilot study and the only star for which H i Lyα was not detected. This non-detection was approximately a factor of 10 below the expected flux level based on an extrapolation of the F(Lyα) versus F(Mg ii) relation (France et al. 2013) and was made more surprising by the solid detection of C ivλ1548 emission. This led us to speculate about the possibility of a high molecular fraction atmosphere that suppresses atomic hydrogen emission, or possibly an "on/off" behavior where the basal flux level is very low and the chromospheric and coronal emission can only be observed during flares (as has been noted for the M8 star VB 10 by Linsky et al. 1995; however, more recent observations demonstrated an "always on" behavior, Hawley & Johns-Krull 2003). Subsequently, X-rays from GJ 1214 were detected by XMM-Newton (Lalitha et al. 2014). All of the source counts during the XMM observations occurred in a single time bin, suggesting that the "on/off" scenario may be correct.

The new MUSCLES observations instead argue for a weak but persistent high-energy flux from GJ 1214. We have detected the full complement of FUV and NUV emission lines from GJ 1214 with the deeper MUSCLES treasury data (including direct observation of Lyα, see Paper II; Table 3). While the Lyα reconstruction is uncertain due to the low S/N and the small fraction of the line profile that is detected, we can compare the observed flux from the star over multiple epochs. The Lyα (STIS G140M) and C iv + Mg ii (COS G160M + STIS G230L) observations in France et al. (2013) were separated by roughly 15 months, while the 2015 MUSCLES Treasury data were obtained over a period of <2 days. We find that the observed flux level from GJ 1214 in 2015, F₂₀₁₅(Lyα) = 1.8 (±0.3) × 10⁻¹⁵ erg cm⁻² s⁻¹, is consistent with the upper limit presented previously, F₂₀₁₃(Lyα) ≤2.4 × 10⁻¹⁵ erg cm⁻² s⁻¹. The reconstructed flux is approximately twice this value, F_2015,recon(Lyα) = 5.5 ×10⁻¹⁵ erg cm⁻² s⁻¹, which is still roughly a factor of three lower than the expected reconstructed Lyα flux based on the F(Lyα) versus F(Mg ii) relation (Youngblood et al. 2016). We find that the total C iv brightness has increased by a factor of two compared to our earlier study (F₂₀₁₅(C iv) = 5.2 (±1.2) × 10⁻¹⁶ erg cm⁻² s⁻¹ versus F₂₀₁₃(C iv) = 2.6 (±0.5) ×10⁻¹⁶ erg cm⁻² s⁻¹), while the Mg ii flux is approximately constant (F₂₀₁₅(Mg ii) = 1.7 (±0.3) × 10⁻¹⁵ erg cm⁻² s⁻¹ versus F₂₀₁₃(Mg ii) = 2.2(±0.2) × 10⁻¹⁶ erg cm⁻² s⁻¹).

Table 3.  MUSCLES Broadband and Emission Line Luminosity^a,b,c

Target	LBol	f(XUV)	f(FUV)	f(NUV)	f(Lyα)	f(N v)	f(C iv)	f(Mg ii)
GJ 1214	31.15	−4.50	−4.45	−4.20	−4.52	−6.48	−6.03	−5.53
GJ 176	32.12	−4.46	−4.40	−3.81	−4.51	−6.61	−6.01	−4.91
GJ 436	32.02	−4.41	−4.52	−4.02	−4.59	−6.93	−6.54	−5.33
GJ 581	31.64	−4.50	−4.84	−4.18	−4.90	−7.23	−6.70	−5.69
GJ 667C	31.70	−4.21	−4.17	−3.86	−4.23	−7.11	−6.47	−5.33
GJ 832	31.79	−4.30	−4.29	−3.60	−4.36	−6.80	−6.46	−5.07
GJ 876	31.69	−4.04	−4.53	−4.33	−4.68	−6.24	−5.91	−5.77
HD 40307	32.98	−4.57	−4.31	−2.41	−4.37	−8.20	−7.21	−5.15
HD 85512	32.78	−4.82	−4.45	−3.01	−4.51	−7.79	−6.89	−4.93
HD 97658	33.12	−4.61	−4.36	−2.05	−4.43	−7.86	−6.97	−5.02
Eri	33.08	−4.15	−4.10	−1.97	−4.20	−6.97	−6.13	−4.30

Notes.

^aFlux measurements are averaged over all exposure times for individual observations, and broadband SEDs are constructed as described in Loyd et al. (2016). ^bAll quantities presented as log₁₀(L_Bol) or f(band) = log₁₀(L(band)/L_Bol). ^cBroadband bandpasses are defined as Bol Δλ = $5\;\mathring{\rm A} -\infty$, XUV Δλ = 5–911 Å, FUV Δλ = 912–1700 Å (including Lyα), NUV Δλ = 1700–3200 Å.

Download table as:  ASCIITypeset image

GJ 1214 displays fractional luminosities that are typical for the mid/late M dwarfs in the MUSCLES survey (Table 3; Figures 10 and 11). It has the lowest absolute levels of high-energy radiation (Figure 8) due to its low mass. The low intrinsic luminosity level and the star's relatively large distance (d = 13 pc) can explain the non-detection of Lyα in the previous work. It appears that GJ 1214 is simply a scaled-down version of the typical planet-hosting mid-M dwarf, and is not in a fundamentally different state of chromospheric activity. While GJ 1214 clearly has flares, we have now shown that the basal flux level of this star is similar to other low-mass exoplanet host stars. Combining this result with the high activity levels on M5 stars such as GJ 876, we conclude that special time-dependent photochemistry (other than the incorporation of impulsive flares and longer-term variability; Section 1) is not necessary down to spectral type M5.

The MUSCLES panchromatic SED creation relies on direct observation at all wavelengths, except in the red/IR wavelengths (λ > 6000 Å) and the XUV/FUV region between 50 and 1170 Å. As discussed in the Introduction, the EUV regulates heating and mass-loss in planetary atmosphere, particularly for short-period planets around stars with large EUV and X-ray fluxes. However, obtaining a complete EUV spectrum of any cool star other than the Sun is currently impossible due  to attenuation by the ISM. Our approach to filling in this observationally inaccessible gap is to employ a coronal model from 50 to 100 Å (a single or two-temperature APEC model, Smith et al. 2001) and models of solar active regions from 100 to 1170 Å (Fontenla et al. 2011; Linsky et al. 2014). This approach assumes a scaling between the chromospheric Lyα emission and the chromospheric + transition region + coronal flux that contributes to the EUV bands, particularly between 100 and 400 Å.

An alternative approach has been taken by Sanz-Forcada et al. (2011), who used X-ray observations of the coronae of exoplanet host stars and an emission measure distribution technique to predict the EUV and part of the FUV spectra of low-mass host stars. The synthetic spectral output of coronal models for three of the MUSCLES stars ( Eri, GJ 436, and GJ 876) are available on the X-exoplanets website,²⁷ and provide overlap with the high S/N MUSCLES data in the 1150–1200 Å region. We downloaded these synthetic spectra to compare the coronal model fluxes with the data for the bright C iiiλ1175 multiplet. These lines are formed in the transition region at T_form ∼ 6 × 10⁴ K, and therefore provide a good diagnostic for how well coronal models are able to reproduce the UV emission throughout the FUV. The X-exoplanets synthetic spectra are provided in units of photons s⁻¹ cm⁻² bin⁻¹ with a variable bin size. The spectra are multiplied by a factor of (bin size)⁻¹ and then integrated over wavelength to compute the integrated C iii line photons s⁻¹ cm⁻². The MUSCLES data were converted to photons and integrated over the same wavelength interval (1174–1177 Å).

We find that the coronal models systematically underpredict the UV emission line strengths by factors of a few to tens (Table 4). For Eri, we find F_data(C iii) = 1.8 × 10⁻² photons s⁻¹ cm⁻² versus F_Xexoplanets(C iii) = 6.6 × 10⁻⁴ photons s⁻¹ cm⁻². For GJ 436, we find F_data(C iii) = 4.2 × 10⁻⁵ photons s⁻¹ cm⁻² versus F_Xexoplanets(C iii) = 8.0 × 10⁻⁶ photons s⁻¹ cm⁻². For GJ 876, we find F_data(C iii) = 6.5 × 10⁻⁴ photons s⁻¹ cm⁻² versus F_Xexoplanets(C iii) = 1.9 × 10⁻⁵ photons s⁻¹ cm⁻². The coronal models underpredict these transition region fluxes by factors of approximately 27, 5, and 33 for Eri, GJ 436, and GJ 876, respectively. This difference underscores the importance of having empirical inputs for transition region and chromospheric emission in the calculations of the EUV flux from cool stars (see also the discussion presented in Linsky et al. 2014). The systematic underprediction of the transition region emission brings into question the accuracy of the coronal models in the 400–900 Å EUV region which is dominated by chromospheric and transition region line and continuum spectra. Another important factor is the time variability in the coronal and chromospheric emission from these stars, which may not be the same. The MUSCLES database provides an excellent resource for emission measure distribution-based stellar atmosphere models that simultaneously take into account both the intermediate- and high-temperature regions of the stellar atmosphere (e.g., Chadney et al. 2015) and may provide more accurate EUV flux estimates for the exoplanet community.

Table 4.  Comparison of MUSCLES C iiiλ1175 Emission and X-exoplanets Spectral Synthesis Prediction

Star	F(C iii), MUSCLES	F(C iii), X-exoplanets a
GJ 876	6.5 × 10−4	1.9 × 10−5
GJ 436	4.2 × 10−5	8.0 × 10−6
Eri	1.8 × 10−2	6.6 × 10−4

Note.

^aAll fluxes in units of (photons s⁻¹ cm⁻²).

Download table as:  ASCIITypeset image

It has been suggested that stellar and exoplanetary magnetic fields can interact in exoplanetary systems (Shkolnik et al. 2003; Lanza 2008), possibly manifesting as enhanced stellar activity relative to nominal age–rotation–activity relationships for isolated main-sequence stars (Barnes 2007; Mamajek & Hillenbrand 2008). The magnitude of this interaction, as measured by the energy dissipated in the stellar atmosphere, may be proportional to the strength of the stellar magnetic field, the planetary magnetic field, and the relative speed of the planet's orbital velocity compared to the stellar magnetic rotation rate. While the stellar magnetic field and orbital velocity can be readily measured, exoplanet magnetic fields have proven notoriously hard to detect (see, e.g., Lecavelier des Etangs et al. 2011 and Hallinan et al. 2013), with very few possible detections from low-frequency radio observations (Lecavelier des Etangs et al. 2013) and early-ingress measurements of NUV and optical transit light curves (Fossati et al. 2010; Lai et al. 2010; Vidotto et al. 2010; Llama et al. 2011; Cauley et al. 2015). If the above dependencies on the magnetic field and velocities are valid, then a general trend of enhanced energy dissipation, which could be observed as enhanced emission from the stellar corona (X-rays; Kashyap et al. 2008) or chromosphere (Ca ii; Shkolnik et al. 2005), should correlate with M_plan/a_plan, where M_plan is the planetary mass and a_plan is the semimajor axis. Because magnetic field strength increases with planetary mass in the solar system, one would expect that the most massive, closest-in planets in exoplanetary systems should produce the largest signal on their host stars. However, almost every claimed detection of a modulation in activity on exoplanet host stars or connection with orbital phase has not been confirmed by follow up observation or re-analysis of the same data set (e.g., Shkolnik et al. 2008; Poppenhaeger et al. 2010).

The search for star–planet interaction (SPI) has predominantly focused on hot Jupiters orbiting F, G, and K stars. The MUSCLES database allows us to examine potential star–planet interactions for a range of planetary masses and semimajor axes on a new class of optically inactive low-mass star. Furthermore, the MUSCLES database allows us to explore SPI as a function of emission line formation temperature, which can constrain the possible location of magnetic field line connection in the stellar atmosphere and the subsequent location of plasma heating. Figure 16 shows a correlation between the N v fractional luminosity and M_plan/a_plan. The N v – M_plan/a_plan relation has a Pearson correlation coefficient of 0.83 and a statistical non-correlation likelihood of 0.035. This suggests that the systems with close-in, massive planets may indeed be generating enhanced transition region activity, as probed by this ∼2 × 10⁵ K gas.²⁸ On the other hand, we do not observe a correlation with the lower-temperature chromospheric gas traced by Mg ii (T_form ∼ (1–2) × 10⁴ K). The Mg ii–M_plan/a_plan relation, for instance, displays a correlation coefficient and probability of a null correlation, −0.3 and 0.06, respectively. Figure 17 presents the Pearson correlation coefficients for the relation between the fractional luminosity and M_plan/a_plan for several lines in the MUSCLES spectra. While the null-hypothesis cannot be ruled out with high confidence for individual hot gas lines (i.e., the n values are not very low), the trend with line formation temperature suggests that the connection between hot gas emission enhancement and M_plan/a_plan may be real. This suggests that the SPI correlation would not be observed in Ca ii, which has a formation temperature distribution comparable to Mg iiin M dwarfs (Fontenla et al. 2016).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The correlation of the hot gas lines in the MUSCLES sample is somewhat surprising given the null result found by Shkolnik (2013) for GALEX fluxes. As the GALEX observations are broader band measurements, there is an uncertain and variable contribution from the stellar photosphere that must be subtracted, lines from a range of formation temperatures are simultaneously included, and the highest temperature lines (e.g., N v) are excluded. Therefore, potential SPI signals are significantly diluted in UV imaging surveys relative to the MUSCLES survey, which focuses on lower-mass stars with less continuum emission and spectrally resolved line profiles that can be robustly measured. This result argues that very low spectral resolution (R < 500) observations at FUV or NUV wavelengths are not a promising avenue for discovering SPI-driven activity enhancements in M dwarfs.

We emphasize that given the small sample size, the detection of SPI in the MUSCLES sample should be viewed with caution. One explanation for these results is a simple scaling with stellar effective temperature. Our inactive K dwarfs have low fractional luminosities and low-mass planetary systems, driving one end of the correlations in some cases. A larger sample that is better controlled for systematics is needed to confirm the results presented here. A future spectroscopic survey of G, K, and M exoplanet host stars with the COS G130M mode (to cover N v, Si iv, and C ii) could be a promising data set with which to characterize enhanced magnetospheric stellar emission generated by SPI.