Detecting Biosignatures in the Atmospheres of Gas Dwarf Planets with the James Webb Space Telescope

For emission spectroscopy the P–T profiles for our targets are adjusted and based on the P–T profile of Earth. According to Seager et al. (2013a) - who approximates a P–T profile for a cold Haber world with a T_eq = 290 K—the precise P–T structure of the atmosphere is less important than photochemistry for a first order description of biosignatures in H₂ rich atmospheres. We utilize public atmospheric data available from Public Domain Aeronautical Software (PDAS) to produce a P–T profile for Earth, ¹¹ and shift the profile based on the planet surface pressure so that 1 bar atmospheres matches the equilibrium temperature of the targets.

We use an isothermal P–T profile for the transmission spectroscopy. Unlike emission spectroscopy, an isothermal P–T profile will not produce a featureless transmission spectrum. However, isothermal profile may be overly simplified and can introduce bias in retrieval analysis (Rocchetto et al. 2016).

We calculate the volume mixing ratio (VMR) using values for a cold Haber world (Seager et al. 2013b) and the VMRs are reported in Tables 2 and 3. We assume the same chemistry of a cold Haber world to focus on the effects of temperature on the magnitude of transmission spectrum. The VMR is calculated using the following equation:

$$\begin{eqnarray}&&\mathrm{VMR}=\displaystyle \frac{{n}_{i}}{{\rm{\Sigma }}{n}_{i}},\end{eqnarray} \tag{ 1 }$$

where n_i is the mixing ratio from (Seager et al. 2013b) for a given species at 1 bar pressure. The species with opacity information in petitRADTRANS are: H₂O, CO₂, CH₄, H₂, CO, HCN, OH, and NH₃. In addition, we include H₂ and He for collision-induced absorption and N₂ as a filler gas.

The VMRs from Seager et al. (2013b) are summed and normalized by dividing by the summation so the total VMR adds to 1.0.

Since petitRADTRANS takes in mass mixing ratio (MMR), we calculate MMR as follows:

$$\begin{eqnarray}&&{\mathrm{MMR}}_{i}=\displaystyle \frac{{\mu }_{i}}{\mu }{\mathrm{VMR}}_{i},\end{eqnarray} \tag{ 2 }$$

where μ_i is the mass of the species in atomic unit, μ is the MMW for the atmosphere in atomic unit, and VMR_i is the volume mixing ratio. The MMW is calculated as

$$\begin{eqnarray}&&\mu =\sum {\mu }_{i}\cdot {\mathrm{VMR}}_{i}.\end{eqnarray} \tag{ 3 }$$

Radius and surface gravity are also the input for petitRADTRANS. The radii are taken from the NASA Exoplanet Archive (Table 1). To calculate surface gravity, we also get masses from NEA, except for the mass of LP 791-18 c for which we use a mass–radius scaling relation (Crossfield et al.2019).

With the inputs that are described in previous sections, we model the emission spectra (f_p ) as shown in Figure 2. The output flux unit for petitRADTRANS in mJy is converted to f_p /f_⋆, which is the input for PandExo (Batalha et al. 2017) to generate simulated JWST data.

Zoom In Zoom Out Reset image size

To calculate stellar flux (f_⋆), we use the PHOENIX model grids (Husser et al. 2013). We resample the wavelength grid of the synthetic planet spectrum from petitRADTRANS onto the wavelength grid of the PHOENIX model to determine f_p /f_⋆. Examples of planet-star contrast spectra are shown in Figure 3.

Zoom In Zoom Out Reset image size

Some examples of the modeled transmission spectra are shown in Figure 4. The reference pressure for all targets is set to P₀ = 1.0 bar and the planet radius and surface gravity values (Table 1) are given at P₀ in petitRADTRANS.

Zoom In Zoom Out Reset image size

We consider the NIRSpec, MIRI, and NIRISS instruments for this work. ¹² A summary of the instrument specifications is provided in Table 4. These instruments allow for a range of wavelengths that cover major NH₃ features near 1.0–1.5, 2.0, 2.3, 3.0, 5.5–6.5, and 10.3–10.8 μm, and spectral features from other molecular species such as CO, CO₂, HCN, and CH₄. NIRCam has lower throughput than NIRSpec, MIRI, and NIRISS so we omit this instrument from this study.

Ammonia has major absorption features at 10.3–10.8 μm, so we test the capabilities of MIRI LRS (Kendrew et al. 2015) to detect this feature. We exclude the use of MIRI MRS because although this mode has a higher resolution, it has a lower throughput than MIRI LRS (Glasse et al. 2015). We note that MIRI MRS has been considered in other non-transiting exoplanet studies (Snellen et al. 2017). The use of MIRI MRS for transiting exoplanets is still possible; capabilities of this mode will be further investigated in Cycle 1 (Kendrew et al. 2018; Deming et al. 2021).

We use PandExo (Batalha et al. 2017) to simulate observations for our targets. We assume a stack of 10 eclipse observations, an 80% detector saturation level, and a floor noise of 40 ppm. The floor noise is set following Chouqar et al. (2020), adopting a value of 40 ppm between the 30 ppm reported in Beichman et al. (2014) and 50 ppm reported in Greene et al. (2016). The noise floor in PandExo is only for one transit, and goes as $1/\sqrt{N}$ for stacking N eclipses.

To further set up a run in PandExo, we input our simulated f_p /f_⋆ spectra and PandExo uses the incorporated PHOENIX model library (Husser et al. 2013) as the stellar input for the simulated observations based on the effective temperature, surface gravity, K-band magnitude, and metallicity [Fe/H] for each target host star.

We define the significance of spectral feature detection using a signal to noise ratio (S/N):

$$\begin{eqnarray}&&{\rm{S}}/{\rm{N}}=\frac{{({R}_{p}/{R}_{\star })}^{2}-\bar{{({R}_{p}/{R}_{\star })}^{2}}}{{\sigma }_{({R}_{p}{/{R}_{\star })}^{2}}},\end{eqnarray} \tag{ 4 }$$

where ${({R}_{p}/{R}_{\star })}^{2}$ is the transmission signal from petitRADTRANS, $\overline{{({R}_{p}/{R}_{\star })}^{2}}$ is median of the transmission signal from petitRADTRANS ¹³ and ${\sigma }_{{({R}_{p}/{R}_{\star })}^{2}}$ is the uncertainty.

Following Wunderlich et al. (2021) and Chouqar et al. (2020) we note ammonia features in the near and mid-infrared have other spectral features that overlap and can obscure these feature (Figure 5). For example, the 2.0 μm NH₃ feature overlaps with H₂0 and H₂–H₂ features. In the mid-infrared the 10.3–10.8 μm NH₃ feature overlaps with H₂-H₂ (Wunderlich et al. 2021).

Zoom In Zoom Out Reset image size

Similar to Chouqar et al. (2020), we neglect the complications from these overlapping features of H₂O, NH₃ and other species. Our detection metric focuses on the S/N for detecting any spectral features, whether or not they are overlapped. The metric will help in determining the number of transits to significantly detect potential biosignatures for given atmospheric compositions. In addition, we show in Section 6 that H₂O and NH₃ can be independently detected and constrained in atmospheric retrieval analyses despite overlapping spectral features.

To simulate JWST transmission spectra, we utilize the Near InfrarRed Spectrograph (NIRSpec) instrument with the G235M and G395M modes which covers the wavelength ranges of 1.7–3.0 μm and 2.9–5.0 μm, respectively. NIRSpec/G395M mode has an expected floor noise of 25 ppm (Kreidberg et al. 2014). We use a fixed number of 10 transits for our simulations with PandExo as with the MIRI LRS simulation.

We assume an optimistic noise floor level for NIRISS (SOSS) and the NIRSpec modes. Greene et al. (2016) notes that "The best HST WFC3 G141 observations of transiting systems to date have noise of the order of 30 ppm (Kreidberg et al. 2014)...". The noise floor in PandExo is only for 1 transit, and goes as $1/\sqrt{N}$ for stacking N transits.

We also consider the use of the Near Infrared Imager and Slitless Spectrograph (NIRISS) instrument in the Single Object Slitless Spectroscopy (SOSS) mode, which covers the 0.8–2.8 μm range. This compliments the wavelength coverage for the NIRSpec/G235M and NIRSpec/G395M modes. An optimistic noise floor of 20 ppm is adopted for NIRISS (SOSS) (Greene et al. 2016; Fortenbach & Dressing 2020).

Pandexo simulations for these instruments for LHS 1140 b, LP 791-18 c, TOI-270 d, LHS 1140 b, K2-18 b, K2-3 c, and GJ 143 b are provided in Figure 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (7 panels)

Tarball of figure set images

We also simulate whether the 6.1 and 10.3–10.8 micron feature of ammonia is detectable with transmission spectroscopy with MIRI LRS, using our detection metric. We assume the same setup for PandExo as with Section 4.2.

The majority of our targets are not feasible for eclipse spectroscopy with MIRI LRS (Figure 3). This is consistent with other studies that achieving the necessary flux contrast for significant detection of molecular features has proved to be very difficult with MIRI (Batalha et al. 2018; Chouqar et al. 2020). LP 791-18 c has the most promising flux contrast ratio but the emission signal is mostly overwhelmed by the photon noise, assuming 10 transits and ∼40 hr of observing time (Figure 7). Similarly TOI-270 c and TOI-270 d are photon limited. LP 791-18 c, TOI-270 c, and TOI-270 d have an S/N of 0.3σ, 0.2σ, and 0.1σ for the 10.3–10.8 NH₃ feature. K2-3 c, LHS 1140 b, K2-18 b, and GJ 143 b are limited by systemic noise as shown in Figure 3.

Zoom In Zoom Out Reset image size

Based on our transmission spectroscopy detection metric (Equation (4)), we find that the NIRSpec/G235M, NIRSpec/G395M, and NIRISS/SOSS modes are best suited to detect ammonia for our targets, with one exception—LP 791-18 c for which NIRSpec PRISM/CLEAR is optimal because this target does not saturate the detector (Table 5).

We compile a ranked list of targets based on transmission spectroscopy simulations for the NIRSpec/G235M, NIRSpec/G395M, NIRISS/SOSS, and MIRI LRS modes with 10 transits. To compute the rank list (Table 6), we use six major absorption features of NH₃ (1.5, 2.0, 2.3, 3.0, 6.1, and 10.3–10.8 μm).

For the 1.5, 2.0, 2.3, 3.0, 6.1, and 10.3–10.8 μm NH₃ features we find the approximate central wavelength and use three data points [one centered on the approximate central wavelength and two adjacent data points] and compute the S/N based on the average of the three data points.

For the final S/N determination, we square each S/N of the NH₃ feature, take the summation, and take the square root of the summation (Equation (5)) to determine the ranking

$$\begin{eqnarray}&&\lt {\rm{S}}/{\rm{N}}\gt =\sqrt{\displaystyle \sum _{i}{\rm{S}}/{{\rm{N}}}_{i}^{2}}\end{eqnarray} \tag{ 5 }$$

where i indicates NH₃ λ_i features at 1.5, 2.0, 2.3, 3.0, 6.1, and 10.3–10.8 μm.

TOI-270 c is best suited for atmospheric studies with JWST given the S/N of detection features for NH₃. On the other hand, GJ 143 b is the least favorable target given that it saturates the NIRSpec/G235M, NIRSpec/G395M, and NIRISS/SOSS modes due to the brightness of the host star (Table 5).

We show a few examples of transmission spectra for TOI-270 c from Figures 8–10. As the S/N scales inversely with MMW, we find that the ideal observing conditions are atmospheres with a low-MMW and clear atmosphere (Figure 8) as opposed to a high-MMW atmosphere that produces a weaker transmission signal and therefore lower S/N (Figure 9).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For MIRI transmission spectroscopy, We find that the 6.1 µm NH₃ is more promising for detection with transmission spectroscopy with MIRI LRS than the 10.3–10.8 µm NH₃ feature (Figure 10) because of increasing noise toward longer wavelengths.

Zoom In Zoom Out Reset image size

We also explore various conditions and atmospheric scenarios which could affect the detection level of ammonia in the atmosphere of our targets in the near-infrared: (1) varying concentration of ammonia in the atmosphere (Section 5.3) (2) varying atmospheric composition (Section 5.4); and (3) varying cloud decks (Section 5.5).

Using TOI-270 c observations with NIRISS and NIRSpec as an example, we vary the amount of ammonia in the atmosphere (Figure 11; Table 7). Following Seager et al. (2013b), the concentration of ammonia in the atmosphere is proportional to the biomass on the surface. A concentration of 11 ppm NH₃ in the atmosphere of a temperature planet ¹⁴ around a weakly active M dwarf is produced by a biomass density of 1 gm⁻². In contrast, a quiet M-type star would need less biomass—1.4 × 10⁻² gm⁻², to produce the same 11 ppm concentration of NH₃ in the atmosphere (Seager et al. 2013a; Seager et al. 2013b).

Zoom In Zoom Out Reset image size

We choose varying ammonia concentrations of 0.4, 4.0, 40, and 400 ppm. A range of 1–10 ppm concentration of NH₃ is reasonable for cold Haber worlds. Our base simulations of detectability assume a 4.0 ppm concentration of ammonia, so we explore the effects of a varying ammonia concentration on the transmission signal on TOI-270 c. Figure 12 shows that a higher concentration of ammonia produces higher signals for detection with the NIRSpec/G235M, NIRSpec/G395M, and NIRISS/SOSS modes (Table 7). In the calculation of NH₃ signal as shown in Figure 12, we measure the transmission signal difference for the varying levels of ammonia in the atmosphere (0.4, 4.0, 40, and 400 ppm) for the 1.5, 2.0, 2.3, and 3.0 μm ammonia transmission features relative to the base transmission signal for a 0-ppm ammonia mixing ratio.

Zoom In Zoom Out Reset image size

Figure 3 in Seager et al. (2013b) shows that vertical mixing can only increase NH₃ values higher up in the atmosphere, making it more detectable, so our simulation is a less optimistic case than the case that considers vertical mixing.

Theoretical studies and observational works based on data from HST have explored the atmospheric compositions for gas dwarf atmospheres (e.g., Elkins-Tanton & Seager 2008; Miller-Ricci et al. 2008; Seager & Deming 2010; Benneke & Seager 2012). Recently, HST observations showed that water vapor was present in the atmosphere, which indicates a thick hydrogen-rich gas envelope for K2-18 b (Benneke et al. 2019b; Madhusudhan et al. 2020), and evidence of a thick atmosphere for LHS 1140 b (Edwards et al. 2021).

To explore the diversity of gas dwarf atmospheres, we follow Chouqar et al. (2020) to consider the following cases: a hydrogen-rich atmosphere (90% H₂ and 10% N₂), a hydrogen-poor atmosphere (1% H₂ and 99% N₂), ¹⁵ and a hydrogen-intermediate atmosphere (75% H₂ and 25% N₂), in order to determine the effects on the detection of ammonia. Figure 13 shows that the strength of ammonia features corresponds to the amount of hydrogen present in the atmosphere. Compared to a hydrogen-rich and hydrogen-intermediate atmosphere, a hydrogen-poor atmosphere has a weaker transmission signal and S/N—based on our transmission detection metric. A more quantitative summary is given in Table 8.

Zoom In Zoom Out Reset image size

One of the major challenges for the search of NH₃ in gas dwarfs is the presence of clouds, which have been shown to mask transmission features (Helling 2019; Barstow 2021). Clouds have been shown to be present in the atmospheres of gas dwarfs (Kreidberg et al. 2014; Benneke et al. 2019b). For example, GJ 1214 b, has shown a flat transmission spectrum due to the effects of high-altitude clouds (Berta et al. 2012; Kreidberg et al. 2014).

We use petitRADTRANS to model the effect of varying levels of cloud deck structures in the atmospheres (1, 0.1, and 0.01 bar). We find that a decreasing cloud deck pressure (i.e., increasing height of clouds) masks the NH₃ atmospheric features to a near flat continuum and affects the detectability of major NH₃ transmission feature (Figure 14). A more quantitative summary is given in Table 9. ¹⁶

Zoom In Zoom Out Reset image size

We employ the use of a contribution function that indicates locations of a cloud deck layer in the atmosphere that transmission features are produced from and where spectral features begin to become muted.

We find that primarily the transmission features are produced at pressures higher than 10⁻² bar in the atmosphere. Therefore, a cloud deck at the pressure level (and below) starts to mute the spectral features (Figure 15).

Zoom In Zoom Out Reset image size

Hazes can impact spectra and produce flatter transmission spectra similar to clouds (Marley et al. 2013; Wunderlich et al. 2021). Our gray cloud treatment results in the same behavior as hazes in the near-infrared. Super Rayleigh slope due to hazes (Ohno & Kawashima 2020) primarily impact the bluer optical wavelengths—so we do not explore the effects in this study.

We use the simulated JWST data for TOI-270 c as the input. To model the simulated data, we use petitRadTRANS (Mollière et al. 2019) with the following free parameters: surface gravity, planet radius, temperature for the isothermal atmosphere, cloud deck pressure, and mass mixing ratios for different species that are being considered. In a Bayesian framework, we use PyMultiNest (Buchner et al. 2014) to sample the posteriors. The priors are listed in Table 10. The likelihood function is $\sum \exp [-{({ \mathcal D }-{ \mathcal M })}^{2}/{{ \mathcal E }}^{2}]$/2, where ${ \mathcal D }$ is data, ${ \mathcal M }$ is model, and ${ \mathcal E }$ is the error term. The input values of the retrieval tests can be found in Table 10. For PyMultiNest, we use 2000 live points.

Table 10. Parameters used in Retrieval, Their Priors, Input, and Retrieved Values

Parameter	Unit	Type	Lower	Upper	Input	Retrieved
						Fixed	Free
Surface gravity (logg)	cgs	Uniform	2.0	5.0	3.0395	${3.20}_{-0.34}^{+0.19}$	${3.12}_{-0.35}^{+0.24}$
Planet radius (RP )	RJupiter	Uniform	0.2	0.5	0.208	${0.20802}_{-0.00013}^{+0.00035}$	${0.20805}_{-0.00017}^{+0.00035}$
Temperature (Tiso))	K	Log-uniform	10	3300	400	${436}_{-47}^{+53}$	${436}_{-38}^{+53}$
Cloud pressure (log(Pcloud))	bar	Log-uniform	−6	6	5.05	fixed	${2.46}_{-2.18}^{+2.38}$
H2O mixing ratio (log(${\mathrm{mr}}_{{{\rm{H}}}_{2}{\rm{O}}}$))	⋯	Log-uniform	−10	0	−5.44	$-{5.39}_{-0.94}^{+0.31}$	$-{5.67}_{-0.99}^{+0.44}$
CO mixing ratio (log(mRCO))	⋯	Log-uniform	−10	0	−8.25	fixed	$-{8.35}_{-1.09}^{+1.33}$
CO2 mixing ratio (log(${\mathrm{mr}}_{{\mathrm{CO}}_{2}}$))	⋯	Log-uniform	−10	0	−7.55	fixed	$-{8.14}_{-0.99}^{+0.66}$
CH4 mixing ratio (log(${\mathrm{mr}}_{{\mathrm{CH}}_{4}}$))	⋯	Log-uniform	−10	0	−6.99	$-{7.20}_{-1.54}^{+0.56}$	$-{7.34}_{-1.31}^{+0.62}$
OH mixing ratio (log(mrOH))	⋯	Log-uniform	−10	0	−14.47	fixed	$-{6.11}_{-2.38}^{+1.29}$
NH3 mixing ratio (log(${\mathrm{mr}}_{{\mathrm{NH}}_{3}}$))	⋯	Log-uniform	−10	0	−4.86	$-{4.76}_{-0.46}^{+0.23}$	$-{4.89}_{-0.51}^{+0.31}$
H2 mixing ratio (log(${\mathrm{mr}}_{{{\rm{H}}}_{2}}$))	⋯	Log-uniform	−10	0	−0.44	$-{0.29}_{-0.42}^{+0.20}$	$-{0.37}_{-0.44}^{+0.26}$
HCN mixing ratio (log(${\mathrm{mr}}_{\mathrm{HCN}}$))	⋯	Log-uniform	−10	0	−8.27	fixed	$-{8.05}_{-1.28}^{+1.38}$
Wavelength shift (Δλ )	μm	Uniform	−0.1	0.1	0.0	fixed	${0.00066}_{-0.00281}^{+0.00198}$

Download table as:  ASCIITypeset image

In this case, we use the simulated data for the cloud-free low-MMW case for TOI-270 c (Figure 8). In the retrieval, we assume the cloud deck at 10⁵ bar, which is well below the pressure range that contributes to the absorption (i.e., 10⁻³−10 bar). This is therefore an a priori cloudy-free case for the retrieval. Furthermore, we fix the abundance for species whose mass mixing ratios are below 5 × 10⁻⁸ for the following reasons. First, these species have low abundances and may not be practically detected given the JWST data quality. This will be shown in the next example. Second, we want to focus on NH₃ and H₂O in this example and check if the two species can be reasonably measured despite overlapping absorption wavelengths regions.

As shown in Figure 16, NH₃ and H₂O can be detected in our retrieval, and their abundances are within 1σ from the input values. The comparison between simulated data points and retrieved spectra shows a good agreement, although the 2σ region for the modeled spectra does not cover the majority of the data points. This is particularly the case for the 2.0 and 2.3 NH₃ features. We attribute this issue to the limitation of our modeling software in generating arbitrary shapes of spectra, but point out that the limitation of modeling spectra can be properly accounted for by adding a Gaussian process component in the retrieval process (e.g., Wang et al. 2020).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We have two ways of quantifying the detection significance. First, given the ∼50,000 posterior samples and that no point falls into the [10⁻¹⁰ to 10⁻⁹] mixing ratio bin, we have a lower limit of 4.2σ detection for NH₃ and H₂O. Second, using the retrieved NH₃ abundance of −4.76 dex and a 1σ error bar of 0.46 dex (see Table 10), the distance to the lower edge of the prior −10 dex is 11.4σ. This is consistent with values reported in Table 6: the quadrature summation of S/N for the 1.5, 2.0, 2.3, and 3.0 NH₃ features is 10.9σ.

We then run a retrieval analysis on the full parameter set that includes (1) the cloud deck pressure; and (2) all minor species with mass mixing ratio lower than 5 × 10⁻⁸. This is to demonstrate that the retrieval works with the full parameter set and returns a reasonable result when compared to the input parameters.

We have an upper limit for the cloud deck pressure at ∼3 bar. This is consistent with the contribution function (Figure 15) which shows that the majority of planet transmission signal comes from atmospheric layers with pressures lower than 3 bar. Note that the prior for the cloud deck pressure covers a range from 10⁻⁶ to 10⁶ bar. The comparison between the data and the retrieved modeled spectra is similar to Figure 16 and we therefore do not include a comparison figure for this case. The corner plot is shown in Figure 17.

Chouqar et al. (2020) performed a comprehensive study on the properties of the TOI-270 system including the detectability of an atmosphere and individual molecules using the NIRISS (SOSS), and NIRSpec/G395M modes for transmission spectroscopy in the near-infrared.

To calculate the total expected S/N and number of transits to detect spectral features so that a <S/N> ≥ 5, the approach presented in Lustig-Yaeger et al. (2019) is utilized. Similar to Lustig-Yaeger et al. (2019), an S/ N scaling relation is developed by running the PandExo JWST noise model across a grid with transits ranging from 1–100. An S/ N is determined based on the difference between the model spectrum and featureless fiducial spectrum. This is the basis of the approach used, but a more detailed description of the approach is available in Lustig-Yaeger et al. (2019) and Chouqar et al.(2020).

We find that for both instrument modes there are discrepancies between our results for TOI-270 c, also seen for TOI-270 d. Similarly to our work, Chouqar et al. (2020) utilize petitRADTRANS to generate the spectra, for a clear hydrogen-rich atmosphere; however, they used an MMW =2.39 (X_H = 0.9), as opposed to our hydrogen-rich atmosphere with an MMW = 4.55. As a result, their simulated 2.0 μm NH₃ feature for TOI-270 c has an ∼200 ppm signal from the baseline and our spectrum of TOI-270 c has an ∼100 ppm signal.

For TOI-270 c with a H-rich atmosphere, they find that the NIRISS (SOSS) instrument requires only one transit to detect ammonia ¹⁷ with an S/N = 18. We attempt to replicate their results using their MMW value, R∼10, and the number of transits set to 1. For the same 2.0 μm feature using NIRISS (SOSS) we find an S/N = 12.5.

They also find that with NIRSpec/G395M for one transit, the ammonia feature (we assume they are referring to the 3.0 μm feature, based on their Figure 5) has an S/N of 5.0. For NIRSpec/G395M we find an S/N = 4.0 for the 3.0 μm feature. Therefore, our NIRSpec/G395M simulation is roughly consistent with the result in Chouqar et al. (2020).

Wunderlich et al. (2021) performed an assessment of the detectability of biosignatures for LHS 1140 b with NIRSpec/PRISM. For the purpose of comparison, we look specifically at the 3.0 μm NH₃ feature they considered for their study.

In their study, an S/N = 5 is utilized to determine whether or not a spectral feature is detectable. The method employed involves the subtraction of the full transmission spectra with all included absorption species from the spectrum excluding contribution from individual species. The S/N determination is based on Wunderlich et al. (2019).

Differing from petitRADTRANS, which is based on a radiative transfer model, the atmosphere of LHS 1140 b from Wunderlich et al. (2021) is built using the radiative-convective photochemistry-climate coupled model, 1D---TERRA. The 1D---TERRA model has inputs of the following: P–T profiles, initial compositions, stellar spectrum, and ion-pair production rates. Detailed schematics of the full model description is shown in Figure 1 of Scheucher et al. (2020).

Compared to our study, Wunderlich et al. (2021) additionally varies the concentration of CH₄ with the low CH₄ scenario assuming a VMR of ∼1 × 10⁻⁶, which is about two orders of magnitude higher than our VMR for CH₄ of ∼2.9 × 10⁻⁸ for our low-MMW 90% H₂ model atmosphere.

They find that with NIRSpec/PRISM—to detect the 3.0 μm NH₃, the required time for a H₂-dominated atmosphere with a low CH₄ scenario, the minimum number of transits would be ∼30 transits, to achieve an S/N of 5. Overall, Wunderlich et al. (2021) find that to detect NH₃ the required time would be between 10–50 transits (∼40–200) hr of observing time assuming non-cloudy conditions. to Wunderlich et al. (2021) we do not consider the NIRSpec/PRISM as this mode saturates the detector for LHS 1140 b (Table 5; Wunderlich et al. 2021); instead we employed the use of NIRSpec/G235M, NIRSpec/G395M, and NIRISS (SOSS) modes. For LHS 1140 b based on our defined detection metric, we find that using NIRISS (SOSS) for the 3.0 μm NH₃, with 10 transits we would find a 3.1σ detection given clear atmosphere conditions in a H-rich atmosphere, which corresponds to ∼60 hr of observing time. For 30 transits (∼180) hr of observing time) we find an S/N ∼ 4.4σ. This is qualitatively consistent with their conclusion that NH₃ can be detected in 10–50 transits, and the detection of NH₃ can be made in the presence of higher concentration of CH₄.

Ammonia has been proposed as a biosignature in hydrogen-dominated atmospheres; however, it is not immune to false positives (Seager et al. 2013b; Catling et al. 2018). Seager et al. (2013b) defined three major factors that can cause ammonia to be produced abiotically in these atmospheres: (1) a rocky world with a surface temperature of ∼820 K ¹⁸ (2) the natural production of NH₃ in the atmospheres of mini-Neptunes, and (3) planets that have outgassed NH₃ during evolution. Seager et al. (2013b) notes that targets have to be evaluated on a case-by-case basis. Additionally, according to the thesis work by Evan Sneed, ¹⁹ another cause of false positives for NH₃ include comet collisions that contain inorganic ammonia ice.

Other factors beyond the S/N of NH₃ for our targets should be considered when prioritizing targets for JWST time. These factors include precise mass and radius measurements.

For example, only K2-18 b (Benneke et al. 2019b) and LHS 1140 b (Dittmann et al. 2017; Lillo-Box et al. 2020) have better than ∼15% and ∼3% precision in mass and radius measurements. This allows for a proper modeling of the planet interior (Lillo-Box et al. 2020) and atmosphere composition (Madhusudhan et al. 2020), which is essential to interpret the results of detection and non-detection.