The Surface UV Environment on Planets Orbiting M Dwarfs: Implications for Prebiotic Chemistry and the Need for Experimental Follow-up

Our radiative transfer formalism is described in detail in Ranjan & Sasselov (2017). Briefly, we partition the atmosphere into 64 homogenous layers of 1 km width and use a two-stream formalism with Gaussian (single) quadrature closure to compute propagation of UV light through the atmosphere (Toon et al. 1989). Two-stream radiative transfer is monochromatic; we partition our spectra into wavelength bins of 4 nm width¹² , and integrate all optical parameters over these bins. For numerical stability, we assign a ceiling on the per-layer single-scattering albedo ${\omega }_{0}$ of $(1-{1}^{-12})$. We take the surface albedo to be 0.2, a representative value for rocky planets and consistent with past 1D modeling for Earth and Mars (Kasting 1991; Segura et al. 2003; Wordsworth et al. 2015). We take the solar zenith angle (SZA) to be 482, corresponding to the insolation-weighted global mean value (Cronin 2014). These parameter values correspond to global mean conditions; variations in surface albedo and SZA can generally be expected to drive changes in band-averaged fluence of approximately one order of magnitude (OOM) (Ranjan & Sasselov 2017). As in Ranjan & Sasselov (2017) and Ranjan et al. (2017a), the fundamental quantity our code calculates is the surface radiance, i.e., the integral of the intensity field at the planet surface for elevations $\gt 0$. This is the relevant radiative quantity for calculating photoreaction rates at planet surfaces, and in the two-stream formalism, this can be calculated as

$$\begin{eqnarray*}&&{I}_{\mathrm{surf}}={F}_{N}^{\downarrow }/{\mu }_{1}+{F}_{N}^{\mathrm{dir}}/{\mu }_{0},\end{eqnarray*}$$

where I_surf is the surface radiance, ${F}_{N}^{\downarrow }$ is the downward diffuse flux at the planet surface, F^dir_N is the direct flux at the planet surface, ${\mu }_{0}=\cos (\mathrm{SZA})$ is the cosine of the SZA, and ${\mu }_{1}=1/\sqrt{3}$ for Gaussian quadrature (Toon et al. 1989).

Following Rugheimer et al. (2015), we take the planetary atmosphere to be cloud-free, with a surface pressure of 1 bar composed of 0.9 bar N₂ and 0.1 bar CO₂. Our results are insensitive to these assumptions on atmospheric state, because CO₂ absorption ($\lt 204\,\mathrm{nm}$) saturates for ${p}_{{\mathrm{CO}}_{2}}\gt 0.072$ bar.¹³ The absorption of most major atmospheric gases (e.g., N₂, H₂O, CH₄) is degenerate with this CO₂ absorption, meaning that surface UV is insensitive to their abundances (Ranjan & Sasselov 2017). Trace gases whose UV absorption is nondegenerate with CO₂ (e.g., O₃, SO₂) do not build up to levels high enough to affect surface UV for atmospheric boundary conditions corresponding to the modern abiotic Earth, according to the photochemical calculations of Rugheimer et al. (2015).¹⁴ Similarly, the absorptions of both CO₂ and H₂O clouds are degenerate with gaseous CO₂ absorption, and for Earthlike global mean cloud optical depths of 4–10 (Stubenrauch et al. 2013), the surface UV environment is insensitive to the presence of clouds (Ranjan et al. 2017a). Finally, because the thermal emission of the atmosphere is negligible at UV wavelengths, the UV surface radiance is insensitive to the atmospheric temperature/pressure profile. Consequently, we approximate the atmosphere by a simple exponential model, with $T={T}_{\mathrm{surf}}=288\,{\rm{K}}$ throughout. We experimented with a more realistic atmospheric model with dry adiabatic evolution in the troposphere and an isothermal stratosphere, and obtained conclusions identical to the exponential atmosphere; consequently, we elected to use the simpler exponential model in this work.

We used empirically measured UV spectra of M dwarfs to derive TOA fluxes to input into our radiative transfer model. Comparatively few such measurements are available because of the low M-dwarf luminosity and high telluric opacity in the UV. We relied primarily on spectra collected by the MUSCLES project, which obtained high-quality spectra of seven M dwarfs with the Hubble Space Telescope (France et al. 2016; Loyd et al. 2016) and combined them with measurements from other instruments and stellar models to aggregate broadband emission spectra; we use all M-dwarf stars in their sample.¹⁵ The MUSCLES team also aggregated HST and XMM-Newton measurements and a PHOENIX stellar model to create a spectrum for Proxima Centauri, which we also use. Segura et al. (2005) aggregated IUE and HST observations to compile a UV spectrum of AD Leo in quiescence; we used this spectrum as well.¹⁶ Because of their low luminosities, no spectra of M dwarfs of stellar type later than M5 are available in the NUV regime relevant to surficial prebiotic chemistry.¹⁷ To obtain coverage of late-type M dwarfs, we use the M8 "active" model of Rugheimer et al. (2015), which is formed by scaling a spectrum of AD Leo using emission lines and concatenating it to a PHOENIX model. This may be taken to correspond to the spectrum of a highly active M8 star.

M dwarfs are known for their frequent and energetic flares (Osten et al. 2016). Few spectral measurements of M-dwarf flares in the prebiotically relevant 150–300 nm regime are available. The exception is AD Leo; the Great Flare of 1985 on this star was measured by space- and ground-based instruments, and Segura et al. (2010) have synthesized these measurements to compile spectra of this flare at different time points from 100 to 444 nm. We use these flare spectra of AD Leo to understand the impact of M-dwarf flares on the surface UV environment and on prebiotic chemistry. We note that the Great Flare on AD Leo may not be representative of all M-dwarf flares, as it was an exceptionally energetic flare (U-band emission of ${10}^{33.8}$ erg, Hawley & Pettersen 1991) on one of the most active known M dwarfs.¹⁸ However, it is the only broad-coverage (spectral and temporal), spectrally resolved measurement of an M-dwarf flare we are aware of, and hence remains the focus of our investigation. Results derived from this flare may be interpreted as a limiting case of the impact of M-dwarf flares on the surface UV environment.

We bin all data to a resolution of 4 nm to eliminate negative fluxes. Such a coarse resolution is acceptable when working with biological action spectra, which have spectral features with widths on the order of $\gtrsim 10\,\mathrm{nm}$ (see, e.g., Setlow 1974; Rontó et al. 2003; Cnossen et al. 2007). Following Segura et al. (2005), we scale all M-dwarf emission spectra to a distance a such that the flux at distance a is equal to the modern solar constant, including the factor of 0.9 correction for the redshifted spectral energy distributions (SEDs) of M dwarfs:

$$\begin{eqnarray*}&&a=\sqrt{\displaystyle \frac{L/{L}_{\odot }}{0.9}}\ast 1\,\mathrm{au}.\end{eqnarray*}$$

Table 1 provides these distances and also summarizes other relevant properties of the M dwarfs in our sample.

Table 1.  Stars Used in this Study and Associated Properties

Star	${T}_{\mathrm{eff}}({\rm{K}})$	Spectral Class	d (pc)	$a(\mathrm{au})$	Reference	Note
GJ 1214	2953	M4.5	14.6	0.064	1, and sources therein
Proxima Centauri	3042	M5.5	1.3	0.042	2, 3, and sources therein	Flare star
GJ 876	3062	M5	4.7	0.12	1, and sources therein
GJ 436	3281	M3.5	10.1	0.17	1, and sources therein
GJ 581	3295	M5	6.2	0.11	1, and sources therein
GJ 667C	3327	M1.5	6.8	0.11	1, and sources therein
AD Leo	3400	M3.5	4.9	0.16	4, 5, 6 and sources therein	Extremely active star
GJ 176	3416	M2.5	9.3	0.19	1, and sources therein
GJ 832	3816	M1.5	5.0	0.14	1, and sources therein

References. (1) Loyd et al. (2016), (2) Anglada-Escudé et al. (2016), (3) Meadows et al. (2016), (4) Shkolnik et al. (2009), (5) Rojas-Ayala et al. (2012), (6) Segura et al. (2005).

Download table as:  ASCIITypeset image

To establish a basis of comparison between M dwarfs and Sunlike stars, we use the model of Claire et al. (2012) to calculate the spectra of the young Sun at 3.9 Ga. We choose this age because it is consistent with available evidence for the origin of life on Earth (see, e.g., Ranjan & Sasselov 2016, Appendix A); our results are insensitive to the choice of solar age, and are unaffected for solar ages from 3.5 to 4.1 Ga.

To quantify the impact of different UV surface radiation environments on prebiotic photochemistry, we compute biologically effective dose rates, which measure the reaction rates of specific prebiotically important photoprocesses (Cockell 1999; Rontó et al. 2003; Rugheimer et al. 2015). Our method is described in detail in Ranjan & Sasselov (2017) and Ranjan et al. (2017a). Briefly, we compute

$$\begin{eqnarray*}&&D=\left({\int }_{{\lambda }_{0}}^{{\lambda }_{1}}d\lambda { \mathcal A }(\lambda ){I}_{\mathrm{surf}}(\lambda )\right).\end{eqnarray*}$$

${I}_{\mathrm{surf}}(\lambda )$ is the UV surface radiance, computed from our model. ${ \mathcal A }(\lambda )$ corresponds to the action spectrum, which parametrizes the wavelength dependence of a given photoprocess; higher values of ${ \mathcal A }$ mean that a higher fraction of the incident photons are being used in the photoprocess. ${\lambda }_{0}-{\lambda }_{1}$ is the wavelength range over which ${ \mathcal A }(\lambda )$ and ${I}_{\mathrm{surf}}(\lambda )$ are defined. Since D is a relative measure of reaction rate, a normalization is required to assign a physical interpretation to its value. In this paper, we report

$$\begin{eqnarray*}&&\overline{D}=D/{D}_{\oplus },\end{eqnarray*}$$

where D_⊕ is the dose rate on 3.9 Ga Earth. $\overline{D}\gt 1$ means the reaction is proceeding faster than it would have on young Earth; $\overline{D}\lt 1$ means the reverse.

We use action spectra corresponding to simple prebiotically relevant photoprocesses to measure the impact of UV light on nascent life. This distinguishes our work from previous works (Cockell 2000a, 2002; Cnossen et al. 2007; Rugheimer et al. 2015), which used action spectra corresponding to DNA damage in modern organisms. Focus on DNA damage is inappropriate for prebiotic chemistry because (1) DNA is not thought to have been the primordial biomolecule, (2) modern organisms have evolved sophisticated methods to deal with environmental stress, including UV exposure, that would not have been available to the first life, and (3) this approach ignores the role of UV light as a eustressor¹⁹ for abiogenesis. We consider two photochemical reactions: a stressor process, and a eustressor process. We also compute the band-integrated NUV radiation, which is a pathway-independent measure of the abundance of prebiotically useful radiation.

For our stressor process, we use the cleavage of the N-glycosidic bond in the RNA monomer uridine monophospate (UMP) by UV light, which irreversibly destroys this key biomolecule. We take the action spectrum as the product of the UMP absorption spectrum (Voet et al. 1963) and the quantum yield curve. We assume a step function form to the quantum yield curve, with value $4.3\times {10}^{-3}$ for $\lambda \leqslant {\lambda }_{0}$ and $2.5\times {10}^{-5}$ for $\lambda \gt {\lambda }_{0}$, and we consider ${\lambda }_{0}$ values of 193, 254, and 230 nm, consistent with the measurements of Gurzadyan & Görner (1994). The absorption spectra of the other RNA monomers are structurally similar to UMP and the molecules share many photochemical properties.²⁰ Therefore, if a UV environment is destructive for UMP, it should be destructive for the other RNA monomers, and hence for abiogenesis in the RNA world hypothesis, as well. As shorthand, we refer to this photoprocess under the assumption that ${\lambda }_{0}={\rm{X}}\,\mathrm{nm}$ by UMP-X.

For our eustressor process, we use the production of solvated electrons from the irradiation of a tricyanocuprate (${\mathrm{CuCN}}_{3}^{2-}$) complex, which was invoked by Ritson & Sutherland (2012) in their pathway for the selective synthesis of the short sugars (glycolaldehyde, gylceraldehyde) that are key for the synthesis of RNA, and which are generally useful in a wide range of reductive prebiotic chemistry. We again take the action spectrum to be the product of the absorption spectrum (Magnani 2015; Ranjan & Sasselov 2016) and the quantum yield curve. Following the hypothesis of Ritson & Sutherland (2012) that photoionization of the cyanocuprate drives solvated electron production, we assume the QY to be characterized by a step function with value 0.06 for $\lambda \leqslant {\lambda }_{0}$ and 0 otherwise, and we consider ${\lambda }_{0}=254$ and 300 nm, consistent with the empirical constraints of Ritson & Sutherland (2012). Laboratory measurements of the spectral quantum yield of this process are forthcoming (Z. Todd et al. 2017, in preparation); preliminary results suggest a step occurring at ∼250 nm. As shorthand, we refer to this photoprocess under the assumption that ${\lambda }_{0}\,=$ Y nm by CuCN3-Y.

Action spectra typically encode information about relative, not absolute, reaction rates. Consequently, they are generally arbitrarily normalized to 1 at some wavelength (see, e.g., Cockell 1999 and Rugheimer et al. 2015). We normalize these spectra to 1 at 190 nm. We were unable to located absorption cross-section data for tricyanocuprate for $\lambda \lt 190\,\mathrm{nm}$ or for UMP for $\lambda \lt 184\,\mathrm{nm};$ for wavelengths below these cutoffs, we padded the absorption spectra with the shortest wavelength data available. Since $\lambda \lt 190\,\mathrm{nm}$ is shielded out by even modest amounts of CO₂ in the atmosphere, this padding only comes into play when considering nonexistent or tenuous atmospheres, as we briefly consider in Section 5. Figure 1 presents the action spectra.

Zoom In Zoom Out Reset image size

The action spectra discussed above correspond to specific prebiotically relevant photochemical processes. In addition, we calculate the band-integrated surface radiance for $\lambda =200\mbox{--}300\,\mathrm{nm}$, which we term the "NUV radiance." This is based on emerging studies of the wavelength dependence of a number of prebiotically important photoprocesses, which suggest that radiation in the 200–300 nm regime can be useful to prebiotic chemistry, while radiation at wavelengths $\lt 200\,\mathrm{nm}$ seems to be solely destructive (Z. Todd et al. 2017, in preparation). We use the NUV radiance as a process-independent measure of the abundance of prebiotically useful UV radiation; the dose rate for the NUV radiance tracks those of the specific prebiotic photoprocesses, as expected for a generalized measure.

Figure 2 presents the TOA fluxes for the M dwarfs in our study and the corresponding surface radiances on prebiotic Earth-analog planets shielded by a 1 bar atmosphere (0.9 bar N₂, 0.1 bar CO₂, matching Rugheimer et al. 2015). As noted by Rugheimer et al. (2015), the surface conditions are defined by a cutoff at 204 nm imposed by atmospheric CO₂, and minimal attenuation at longer wavelengths. Notably, young-Earth-analog planets orbiting M-dwarf planets are exposed to far less UV radiation than those orbiting Sunlike stars because of the cooler photosphere and hence lower NUV emission of M dwarfs. At short wavelengths, M dwarfs emit proportionately more radiation than the young Sun, but fluence at these wavelengths is robustly blocked by a range of atmospheric absorbers, including CO₂ and H₂O, which shield out $\lt 200\,\mathrm{nm}$ radiation. Consequently, M-dwarf planets, so long as they can retain their atmospheres, are low-UV environments.

Zoom In Zoom Out Reset image size

We quantify this observation by computing the relative dose rates $\overline{D}$ for each of the stars in our sample. These dose rates are presented in Figure 3. The stars are ordered by decreasing T_eff. We observe that with the exception of the exceptionally active star AD Leo and the Rugheimer et al. (2015) M8V "Active" model, which is a scaled AD Leo, all dose rates are suppressed by $\gt 2$ orders of magnitude relative to the young Earth. Putative UV-dependent prebiotic chemistry will proceed at rates 2–4 orders of magnitude slower on planets orbiting non-active M dwarfs compared to the young Earth. Active stars like AD Leo emit more UV radiation, so bioactive fluence on planets orbiting such active stars will only be suppressed by 1–2 orders of magnitude.

Zoom In Zoom Out Reset image size

We considered the hypothesis that different host stars might affect the stressor and eustressor pathways in different ways. Specifically, we considered the possibility that the eustressor reaction rates might fall off more slowly or faster than the stressor reaction rates due to variations in the shape of the SEDs between the different host stars, which would imply more or less favorable venues for abiogenesis, respectively. To test this hypothesis, we calculated ${\overline{D}}_{\mathrm{UMP} \mbox{-} X}/{\overline{D}}_{\mathrm{CuCN}3 \mbox{-} Y}$ for all X and Y. If this ratio is $\lt 1$, the stressor pathway is disfavored over the eustressor pathway relative to Earth, and the environment compares favorably to Earth as a venue for abiogenesis; if this ratio is $\gt 1$, the reverse is true. Figure 4 presents these calculations. In aggregate, the dose rate ratios do not vary much as a function of host star, remaining within an OOM of unity for all stars in this study, and no clear aggregate trends are visible. We attribute this to the fact that the SEDs of the M-dwarf stars considered here are broadly similar in shape to the Sun's SED at the NUV wavelengths that reach the planet surface. We conclude that M-dwarf planets are comparable to early Earth in terms of how much their UV environments favor stressor processes over eustressor processes.

Zoom In Zoom Out Reset image size

Figure 5 presents the TOA fluxes for AD Leo in quiescence and at the peak of the Great Flare (912 s, Segura et al. 2010), and the corresponding surface radiances on a young-Earth-analog planet orbiting it. In flare, the UV output of AD Leo rises by orders of magnitude, and the increase in fluence is not confined to line emission as it was for the young Sun in the study of Cnossen et al. (2007). Rather, the increase is across the UV spectrum, including the continuum.

Zoom In Zoom Out Reset image size

This increase in fluence drives a dramatic increase in reaction rates. Figure 6 shows the dose rates for a young-Earth analog orbiting AD Leo in and out of flare, compared to the young Earth. While in quiescence prebiotically relevant reaction rates are suppressed by 1–2 orders of magnitude relative to the young Earth, in flare the reaction rates are enhanced, by up to 1 OOM relative to young Earth. The increase in UV fluence is relatively spectrally flat across the NUV regime that penetrates the atmosphere to the planet surface, and the flare does not deliver radiation that particularly favors or disfavors stressor processes versus eustressor processes relative to quiescence.

Zoom In Zoom Out Reset image size

Prebiotic chemistry on planets orbiting M dwarfs has access to orders of magnitude less UV radiation than prebiotic chemistry on planets orbiting solar-type stars because the NUV emission by the cooler M dwarfs is lower. This poses a challenge for some origin-of-life scenarios on such worlds. As discussed in Section 1, UV light is a key requirement of several proposed prebotic pathways, including the only known pathways for prebiotic syntheses of ribonucleotides (Powner et al. 2009; Ritson & Sutherland 2012; Patel et al. 2015; Xu et al. 2016). It is possible that these pathways will still function under a low-UV regime, albeit at lower rates. However, it is equally possible that they will not, because of interference or competition with thermal reactions. For example, the core of the sugar synthesis pathway of Ritson & Sutherland (2012) and Patel et al. (2015) involves the reduction of glycolonitrile to glycolaldehyde imine by a photochemically produced solvated electron, which then hydrolyzes to give the desired glycolaldehyde sugar. However, glycolonitrile can also react with H₂S²¹ to produce its alpha-hydroxy thioamide form, which subsequently hydrolyzes to its alpha-hydroxy acid form. If the UV irradiance is low enough, the latter pathway will dominate the former, and the sugar synthesis pathway will not proceed.²² Similarly, the efficacy of UV light at selecting and amplifying the population of biogenic ribonucleotides and generating uridine from cytidine in Powner et al. (2009) will be functions of the irradiation level. These examples illustrate the need to characterize the sensitivity of putative prebiotic pathways to the amplitude of irradiation, to understand whether they can proceed under the lower-UV conditions on planets orbiting M dwarfs. In sum, it is unclear whether UV-dependent prebiotic pathways that may have been important to the origin of life on Earth can proceed in the low-UV environment on planets orbiting M dwarfs, such as the recently discovered habitable-zone planets Proxima Centauri b (Anglada-Escudé et al. 2016), LHS 1140b (Dittmann et al. 2017), or the habitable-zone planets orbiting TRAPPIST-1 (Gillon et al. 2017).

Even if these pathways do proceed at lower irradiance levels, their reaction rates will likely be orders of magnitude lower than on planets around Sunlike stars. If UV-dependent pathways like the RNA monomer synthesis pathways of Patel et al. (2015) were a rate-limiting step in the origin of life on Earth, then lower M-dwarf NUV radiation could delay the origin of life on planets orbiting M stars by orders of magnitude. If abiogenesis is fast, this is not a problem. However, if abiogenesis is slow, this could pose a challenge. For example, if abiogenesis took 100 Ma on Earth and UV photochemistry was the rate-limiting step, then it would take $\gtrsim 10$ Ga around M dwarfs, meaning that only old M dwarfs could host planets with native life. In the worst-case scenario, the Universe might be too young for M-dwarf-orbiting life to have evolved at all.

The UV-scarcity problem may be exacerbated by atmospheric photochemistry, depending on planet history and surface conditions. Luger & Barnes (2015) show that terrestrial planets with significant initial water inventories orbiting in the current habitable zones of main-sequence M dwarfs may have undergone a runaway greenhouse and suffered massive loss of water due to enhanced M-dwarf luminosity in their extended pre-main sequence phase, which could have generated high levels of UV-shielding O₂ and O₃ if not balanced by loss to the mantle and crust. Similarly, Harman et al. (2015) show that Earthlike planets with significant CO₂ inventories²³ orbiting M dwarfs can build up abiotic ozone layers comparable in strength to modern Earth's if sinks of O₂ and CO are small, because the greater FUV/NUV ratio of M dwarfs means that O₃ production rates are higher and dissociation rates are lower. Gao et al. (2015) demonstrate that CO₂-rich, H₂O-poor planets orbiting M dwarfs can also build up abiotic ozone layers comparable to modern Earth's, and for high H levels can also build up H₂O₂ , which can also act as a UV shield at elevated levels. Wordsworth & Pierrehumbert (2014) showed that water-rich planets with low inventories of non-condensible gases like N₂ may also abiotically oxygenate. In sum, planets orbiting M dwarfs with inefficient sinks of O₂ and CO, with desiccated CO₂ rich atmospheres, or lacking non-condensible background gases, may photochemically produce atmospheric UV shields like O₃ that can further block the already low M-dwarf UV from the planet surface.

One might argue that our focus on UV light is unwarranted because it is not certain that UV light played a role in the origin of life on Earth, let alone on other planets; indeed, UV light would be altogether absent in the deep-sea hydrothermal vent hypothesis for the origin of life (e.g., Martin et al. 2008). Our rationale for focusing on UV light is threefold. First, the nucleotides of RNA show evidence of selection pressure from UV irradiation, suggesting they arose in a UV-rich environment (Rios & Tor 2013; Beckstead et al. 2016; Pollum et al. 2016). If one assumes that RNA were the primordial autocatalytic replicator, as is the paradigm in the RNA world hypothesis (Gilbert 1986), this implies that UV light was abundant during abiogenesis on Earth. Second, only UV light has been empirically demonstrated to drive the only known prebiotically plausible pathways for the origin of the pyrimidine RNA monomers and for the selective sugar synthesis pathway that sidesteps the asphaltization problem (Benner et al. 2012; Patel et al. 2015). It is possible that a different energy source could substitute for UV, but this has not yet been demonstrated in the laboratory. Third, it is difficult to envision sources of free energy that can substitute for UV light in general. The energetic nature of UV light means that it is capable of directly affecting or altering molecular electronic structure, permitting it to effect irreversible changes in the entropic states of molecular systems. To effect similar changes thermally requires sources with temperatures on the orders of thousands of degrees Kelvin, coupled to quenching mechanisms to prevent thermal relaxation to the equilibrium state (Pascal 2012). Two such high-temperature planetary sources are lightning and impacts, which have been contemplated as energy sources for prebiotic molecular synthesis (Chyba & Sagan 1992; Nava-Sedeño et al. 2016). However, these sources face the problem of abundance: lightning and impact-driven shockwaves are estimated to have delivered three orders of magnitude less energy than UV photons ($\lt 300\,\mathrm{nm}$) to the young Earth (Ferris & Chen 1975). Hence, it is difficult to avoid the conclusion that a paucity of UV light might pose a problem for abiogenesis scenarios on M-dwarf planets like Proxima Centauri b, LHS 1140b, or the TRAPPIST-1 planets. Conversely, if life is found on an M-dwarf planet, it might imply a pathway to the origin of life that is very different from what seems to have played out on Earth.

We considered whether the potential UV-paucity problem could be solved with thinner atmospheres that block less UV light. We might imagine that in some regions of parameter space, elevated M-dwarf EUV emission or interactions with a more intense M-dwarf stellar wind could at least transiently strip the atmosphere from habitable-zone planets, similar to the scenario Tian et al. (2009) calculated for the young Mars due to elevated early solar EUV emission, or the scenario of Dong et al. (2017) for the planets of the TRAPPIST-1 system due to enhanced stellar wind interactions. We repeated our study for a tenuous atmosphere of 1 microbar N₂–CO₂ that provided essentially no attenuation of incoming fluence. Relative dose rates did rise by 1–2 orders of magnitude. However, the strongest increases in reaction rates were for destructive reactions (i.e., UMP photolysis), which is unsurprising: the new fluence admitted by the thinner atmosphere were the destructive far-UV (FUV) ($\lt 200\,\mathrm{nm}$) wavelengths, making the environment less clement for abiogenesis. In addition, such low atmospheric pressures pose other problems for prebiotic chemistry: for example, liquid water is not stable at pressures below the triple point, i.e., ${P}_{0}\lt 6\times {10}^{-3}$ bar (Lide 2009), meaning aqueous-phase prebiotic chemistry would face challenges on an airless world. We conclude that atmospheric stripping cannot solve the UV-paucity problem for prebiotic chemistry.

Our analysis has focused on main-sequence M dwarfs. Young M dwarfs, especially pre-main-sequence M dwarfs, emit a larger fraction of their bolometric luminosity as NUV radiation than in their main-sequence phase (Shkolnik & Barman 2014). The lowest-mass M dwarfs remain in this state for timescales of up to ∼1 Gyr, and one might speculate whether planets orbiting such stars might receive NUV irradiation that is more similar to planets orbiting Sunlike stars. However, the bolometric luminosity of such stars is also higher; planets in the main-sequence habitable zones of such stars are liable to be in a runaway greenhouse state during their pre-main-sequence evolution, with temperatures globally above the boiling point of water and possibly as high as $\gtrsim 1000\,{\rm{K}}$ (Ramirez & Kaltenegger 2014; Luger et al. 2015; Luger & Barnes 2015; Schaefer et al. 2016). Hence, these objects will lack clement conditions for aqueous-phase prebiotic chemistry, and are unlikely to be habitable during this phase. Planets orbiting farther out, in the pre-main-sequence habitable zone, may remain habitable for up to gigayear timescales (Ramirez & Kaltenegger 2014), and will on average experience NUV irradiation (175–275 nm) up to 1 OOM higher than planets orbiting at equivalent bolometric instellations during the star's main-sequence phase (Shkolnik & Barman 2014). However, these objects will exit the habitable zone when their host star joins the main sequence and reduce their bolometric luminosity. Unless global temperatures are subsequently elevated by mechanisms not considered in the traditional Earth-analog habitable-zone calculation, e.g., greenhouse warming due to elevated levels of atmospheric H₂ (Stevenson 1999; Pierrehumbert & Gaidos 2011; Wordsworth & Pierrehumbert 2013; Wordsworth et al. 2017; Ramirez & Kaltenegger 2017), liquid water will freeze at the planet surface and life will be confined to volcanic or subsurface reservoirs, whose ability to produce global atmospheric biosignatures is unclear. More critically, it is uncertain whether a 1-OOM enhancement²⁴ in 175–275 nm fluence is adequate to compensate for the 2–4 OOM suppression in bioactive NUV fluence on M-dwarf planets compared to the early Earth; experimental characterization of the fluence dependence of UV-sensitive putative prebiotic chemistry remains necessary to answer this question. Nevertheless, planets orbiting in the habitable zones of late-type (low-mass) pre-main-sequence M dwarfs remain of special interest from the perspective of UV-sensitive prebiotic chemistry.

We considered whether M-dwarf flares, which are typically considered barriers to habitability (Segura et al. 2010), might provide a partial solution to the problem of UV paucity on M-dwarf planets (Buccino et al. 2007). During the Great Flare of 1985, the NUV output of AD Leo increased dramatically, delivering more NUV radiation to the surface of a young-Earth-analog planet than the young Sun for a period of at least 0.7 hr and possibly longer.²⁵ Flare frequency studies of AD Leo suggest that flares of this strength occur witha  frequency of 0.1 day⁻¹ Pettersen et al. (1984).²⁶ For comparison, the typical irradiation time used in laboratory studies of UV-sensitive prebiotic pathways like C-to-U conversion and glycolaldehyde/glyceraldehyde formation at fluence levels comparable with the young Sun is on the order of 4 hr (Z. Todd et al. 2017, in preparation), a duration comparable to strong M-dwarf flares. One might imagine a scenario whereby photosensitive prebiotic chemistry proceeds during the high-UV flares and ceases during stellar quiescence, providing an activity-powered analog to the terrestrial day/night cycle that would be absent on tidally locked habitable-zone M-dwarf planets. Experimental studies are required to constrain whether shorter duration but higher intensity irradiation corresponding to an M-dwarf flare is sufficient to power putative prebiotic chemistry. Theoretical studies are also required to understand whether the enhanced UV radiation these flares provide also strips the atmosphere, which would negatively impact planetary habitability and obviate this solution. We note that this mechanism could best solve the UV-paucity problem for planets orbiting the most active M stars, like AD Leo. Planets orbiting active stars would experience much lower flare rates; for example, extrapolation of white-light flare frequency studies for Proxima Centauri suggest that Proxima Centauri b would experience UV instellation comparable the Great Flare of AD Leo only $\sim 8\,{\mathrm{yr}}^{-1},5\times$ less frequently than an Earth-analog planet orbiting AD Leo (Davenport et al. 2016).

Our work suggests a critical need for laboratory studies to determine whether putative UV-dependent prebiotic pathways (e.g., Powner et al. 2009; Ritson & Sutherland 2012; Patel et al. 2015; Xu et al. 2016) could function at the $10\mbox{--}1000\times$ lower-UV irradiation accessible on planets orbiting M dwarfs compared to the young Earth. First, the fluence dependence of these pathways should be determined: how do their rates vary as a function of irradiation level, and at which minimum fluence do they shut down? These questions should be probed down to fluence levels as low as ${10}^{-3}\mbox{--}{10}^{-4}$ of that accessible on early Earth, corresponding to the fluence available on quiet M dwarfs. Integrated from 200 to 300 nm, prebiotic chemistry on early Earth would have been exposed to ∼6 × 10¹⁴ photons s⁻¹ cm⁻², implying the need to probe down to photon fluxes as low as ∼6 × 10¹¹ photons s⁻¹ cm⁻² (5 erg s⁻¹ cm⁻² at the 254 nm line generally used in prebiotic chemistry studies).

If such studies determine that one or more of these pathways fail at M-dwarf irradiation levels, experiments should be made to determine whether transient high-UV irradiation due to flares can substitute for steady-state irradiation. The best-established case to simulate is the Great Flare on AD Leo; this event delivered 2 × 10¹⁵ photons s⁻¹ cm⁻² (2 × 10⁴ erg s⁻¹ cm⁻² at 254 nm) to the surface of an orbiting young-Earth analog, integrated from 200 to 300 nm. Such fluence levels would have been available for periods of 0.7–4 hr at intervals of $\approx 240$ hr according to the power law presented in Pettersen et al. (1984). If such intense flares are too infrequent to power prebiotic chemistry, then the influence of lower-intensity, higher-frequency flares may be considered. For example, flares 1/10 as energetic as the Great Flare occur every 60 hr (see Section 6). If flare frequencies $\sim 5\times$ lower than that of AD Leo are found to be sufficient to power UV-dependent prebiotic chemistry, then flare instellation could solve the potential UV-paucity problem for Proxima Centauri b.