Nitrogen Dioxide Pollution as a Signature of Extraterrestrial Technology

Over the last 25 years, more than 4000 exoplanets have been discovered⁴ from both ground- and space-based surveys. We are now entering an era of exoplanet atmospheric characterization, with the soon-to-be-launched James Webb Space Telescope (JWST), Atmospheric Remote-sensing Infrared Exoplanet Large-survey (ARIEL) space telescope, and large ground-based observatories such as the European Extremely Large Telescope (E-ELT), the Thirty Meter Telescope (TMT), and the Giant Magellan Telescope (GMT). The first detection of an exoplanet atmosphere was on a gas giant planet, HD 209458b, in 2001 (Charbonneau et al. 2002). Since then, atmospheres have been detected on exoplanets spanning a wide range of planetary parameter space, and observers are continuing to push the limits toward smaller worlds (Benneke et al. 2019; Tsiaras et al. 2019). The ongoing discovery of exoplanet atmospheres has raised the prospect of eventually identifying potentially habitable planets, as well as the possibility of finding one that may also be inhabited. As a result, the characterization and detection of "biosignatures,"—remote observations of atmospheric spectral features that could potentially indicate signs of life on an exoplanet—has received recent attention as an area of priority for astrobiology⁵ (Seager et al. 2012; Grenfell 2017; Kaltenegger 2017; Catling et al. 2018; Fujii et al. 2018; Meadows et al. 2018; Schwieterman et al. 2018; Walker et al. 2018; Lammer et al. 2019; O'Malley-James & Kaltenegger 2019).

Similar to biosignatures, "technosignatures" refer to any observational manifestations of extraterrestrial technology that could be detected or inferred through astronomical searches. As discussed in the 2018 NASA technosignatures workshop report (Technosignatures Workshop Participants 2018): "Searches for technosignatures are logically continuous with the search for biosignatures as part of astrobiology. As with biosignatures, one must proceed by hypothesizing a class of detectable technosignatures, motivated by life on Earth, and then designing a search for that technosignature considering both its detectability and its uniqueness." Although the science of atmospheric technosignatures is less developed than atmospheric biosignatures, a wide class of possible technosignatures have been suggested in the literature that include waste heat (Dyson 1960; Wright et al. 2014; Kuhn & Berdyugina 2015; Carrigan 2009), artificial illumination (Schneider 2010; Loeb & Turner 2012; Kipping & Teachey 2016), artificial atmospheric constituents (Schneider 2010; Lin et al. 2014; Stevens et al. 2016), artificial surface constituents (Lingam & Loeb 2017), stellar "pollution" (Shklovskii & Sagan 1966; Whitmire & Wright 1980; Stevens et al. 2016), non-terrestrial artifacts (Bracewell 1960; Freitas & Valdes 1980; Rose & Wright 2004; Haqq-Misra & Kopparapu 2012), and megastructures (Dyson 1960; Arnold 2005; Forgan 2013; Wright et al. 2016). This breadth of topics reflects the scope of possibilities for detecting plausible technosignatures, although the sophistication of technosignature science remains in its infancy compared to the rapidly evolving field of biosignatures (Wright 2019; Haqq-Misra et al. 2020).

The history of life on Earth provides a starting point in the search for biosignatures on exoplanets (Krissansen-Totton et al. 2018; Pallé 2018), with the various stages of Earth's evolution through the Hadean (4.6–4 Gyr), Archean (4–2.5 Gyr), Proterozoic (2.5–0.54 Gyr), and Phanerozoic (0.54 Gyr–present) eons representing atmospheric compositions to use as examples of spectral signatures of an inhabited planet. The use of Earth's history as an example does not imply that these biosignatures will necessarily be the most prevalent in the Galaxy, but instead this approach simply represents a place to begin based on the one known example of life. By extension, the search for technosignatures likewise can consider Earth's evolution into the Anthropocene epoch (Crutzen 2006; Lewis & Maslin 2015; Frank et al. 2017) as a template for future observing campaigns that seek to detect evidence of extraterrestrial technology. For instance, Lin et al. (2014) discussed the possibility of detecting signatures of tetrafluoromethane (CF₄) and trichlorofluoromethane (CCl₃F) in the atmospheres of transiting Earth-like planets around white dwarfs with JWST, which could be detectable if these compounds are present at 10 times the present Earth level. These chlorofluorocarbons (CFCs) are produced by industrial processes on Earth, so their detection in an exoplanet atmosphere could be strong evidence for the presence of extraterrestrial technology. This approach does not insist that CFCs or other industrial gases found on Earth will necessarily be the most prevalent technosignature in the Galaxy, but it represents a place to begin defining observables and plausible concepts for technosignatures based upon the one known example of technological civilization.

In this study, we explore the possibility of NO₂ as an atmospheric technosignature. Some NO₂ on Earth is produced as a byproduct of combustion, which suggests the possibility of scenarios in which larger-scale production of NO₂ is sustained by more advanced technology on another planet. Detecting a high level of NO₂ above that of non-technological emissions found on Earth could be a sign that the planet may host active industrial processes. In Section 2, we describe the production reactions of NO₂ and use a one-dimensional photochemical model to obtain self-consistent mixing ratio profiles of nitrogen oxide compounds, on a planet orbiting a Sun-like star, a K6V spectral type (T_eff = 4600 K), and the two M-dwarf stars, AD Leo (T_eff = 3390 K) and Proxima Centauri (T_eff = 3000 K). Using these photochemical results, in Section 3 we calculate the observability of strongest NO₂ features between 0.2 and 0.7 μm and between 1 and 10 μm using a spectral generation model to produce geometric albedo and transit spectra of planets with various facilities such as LUVOIR-15 m, JWST, and the Origins Space Telescope. We discuss the implications of these observations in Section 4 and conclude in Section 5.

Nitrogen oxides (NO_x = NO + NO₂) are among the main pollutants in industrialized locations on the globe. The non-anthropogenic pathways for the production of NO_x can be either emission from soils and wildfires or via lightning in the troposphere.⁶ The primary biogenic source of NO_x is bacteria in soil through nitrification (i.e., bacteria converting ammonia to nitrite and nitrate compounds) or denitrification (process of reducing nitrate and nitrite to gaseous forms of nitrogen such as N₂ or N₂O). The estimated worldwide biogenic and lightning emissions of NO_x compounds are ∼10.6 teragram per year as N (Tg(N) yr⁻¹) (Table 1, Holmes et al. 2013). Lightning contributes about 5 Tg(N) yr⁻¹, which translates to 6 × 10⁸ NO molecules cm⁻² s⁻¹ (Harman et al. 2018).

On the other hand, NO_x compounds are also emitted from anthropogenic sources of combustion processes such as vehicle emissions and fossil-fueled power plants. The role of this industrial production was noted during the Covid-19 pandemic, when global concentrations of NO₂ were observed to decrease between 20% and 40% over urban areas (Bauwens et al. 2020). Indeed, these emissions dominate the production of NO_x compounds in the troposphere more than the biogenic sources, with an estimated rate of 32 Tg(N) yr⁻¹ (Holmes et al. 2013). NO₂ poses harmful health effects that could cause impairment of lung function and respiratory problems (Faustini et al. 2014). Typical concentrations of NO₂ range from 0.01 ppb (parts per billion) to ∼5 ppb depending upon the urbanization, with the higher number correlating to urban areas (Lamsal et al. 2013). The presence of NO_x in the lower troposphere leads to a complex chemistry that results in the formation of ozone (O₃), which is a harmful pollutant in the troposphere and a greenhouse gas. NO_x mixing ratios in excess of 10⁻⁷ would cause severe damage to the O₃ layer and could result in either a climatic warming or cooling, depending upon the amount of NO₂ present (Kasting & Ackerman 1985).

The sinks and sources for NO₂ in the troposphere (≤20 km) are governed by the following reactions. NO₂ photolysis is dominant in the wavelength range 290–420 nm (see Kraus & Hofzumahaus 1998, Figure 2). The lower limit is set by the available solar UV intensity and the upper wavelength limit is determined by the fall-off in the photodissociation cross section. This NO₂ photolysis produces ground-state atomic oxygen, O(³P), along with NO:

$$\begin{eqnarray}&&{{\rm{NO}}}_{2}+h\nu (\lt 420\,{\rm{nm}})\to {\rm{O}}({}^{3}{\rm{P}})+{\rm{NO}}.\end{eqnarray} \tag{ 1 }$$

The O(³P) then can combine with an oxygen molecule to form ozone,

$$\begin{eqnarray}&&{\rm{O}}{(}^{3}{\rm{P}})+{{\rm{O}}}_{2}+{\rm{M}}\to {{\rm{O}}}_{3}+{\rm{M}},\end{eqnarray} \tag{ 2 }$$

which gets destroyed by reoxidizing nitric oxide to nitrogen dioxide:

$$\begin{eqnarray}&&\mathrm{NO}+{{\rm{O}}}_{3}\to {\mathrm{NO}}_{2}+{{\rm{O}}}_{2}.\end{eqnarray} \tag{ 3 }$$

NO also reacts with atomic oxygen (O) and the hydroperoxy radical (HO₂) to generate NO₂:

$$\begin{eqnarray}&&\mathrm{NO}+{\rm{O}}+{\rm{M}}\to {\mathrm{NO}}_{2}+{\rm{M}}\end{eqnarray} \tag{ 4 }$$

$$\begin{eqnarray}&&\mathrm{NO}+{\mathrm{HO}}_{2}\to {\mathrm{NO}}_{2}+\mathrm{OH}.\end{eqnarray} \tag{ 5 }$$

However, these production mechanisms of NO₂ are counteracted when NO₂ reacts again with atomic oxygen to recreate NO:

$$\begin{eqnarray}&&{\mathrm{NO}}_{2}+{\rm{O}}\to \mathrm{NO}+{{\rm{O}}}_{2}.\end{eqnarray} \tag{ 6 }$$

The above reactions just cycle between NO and NO₂, so NO_x is conserved. However, NO₂ also reacts with the OH radical to form nitric acid (HNO₃), which eventually is removed from the atmosphere by rainout, which is a loss process for NO_x:

$$\begin{eqnarray}&&{\mathrm{NO}}_{2}+\mathrm{OH}+{\rm{M}}\to {\mathrm{HNO}}_{3}+{\rm{M}}.\end{eqnarray} \tag{ 7 }$$

Reactions (3) and (6) form a catalytic cycle to destroy ozone with the net reaction

$$\begin{eqnarray}&&{{\rm{O}}}_{3}+{\rm{O}}\to 2{{\rm{O}}}_{2}\end{eqnarray} \tag{ 8 }$$

indicating that high stratospheric NO_x can lead to ozone depletion.

To study the steady-state abundances of NO_x compounds in Earth-like atmospheres, we used a 1D photochemical model (described in Arney et al. 2016; Arney 2019), which is part of a coupled climate–photochemistry model called "Atmos."⁷ The photochemical model is originally based on the one described in Kasting et al. (1979) and has been updated extensively over the years and applied to various planetary and exoplanetary conditions (e.g., Segura et al. 2005; Kopparapu et al. 2012; Domagal-Goldman et al. 2014; Harman et al. 2015, 2018; Lincowski et al. 2018). The model version used here has been updated to correct the deficiencies identified in Ranjan et al. (2020), and the public version of the model is planned to be updated. This model solves a set of nonlinear, coupled ordinary differential equations for the mixing ratios of all species at all heights using the reverse Euler method. The method is first order in time and uses second-order centered finite differences in space. The vertical grid has 200 altitude levels, ranging from 0 km (lower boundary) to 100 km (upper boundary). The version used here includes updates described in Lincowski et al. (2018) and includes 72 chemical species involved in 309 reactions to represent a modern Earth-like planet. We considered a Sun-like star, a K6V stellar spectral type, and two M-stars (AD Leo and Proxima Centauri) in this study. For the Sun-like star we used the model of Chance & Kurucz (2010); for the K6V star, we used the spectrum of HD 85512 from the Measurements of the Ultraviolet Spectral Characteristics of Low-mass Exoplanetary Systems (MUSCLES) treasury survey (France et al. 2016; Loyd et al. 2016; Youngblood et al. 2016); for AD Leo and Proxima Centauri, we used stellar spectra described in Segura et al. (2005) and Meadows et al. (2018), respectively. Planets around the other stars are placed at the Earth-equivalent flux distance.

For each Earth-like planet around its host star, we ran the model to steady state to obtain the mixing ratio profiles of all gaseous species, including NO₂. We have used a surface NO₂ molecular flux of 8.64 × 10⁹ molecules cm⁻² s⁻¹ as the standard Earth-level (1×) flux in our simulations. This number comes from converting the estimated rate of 32 Tg(N) yr⁻¹ anthropogenic NO_x compound emissions in the troposphere⁸ to the molecular flux. We also include a fixed biogenic flux of NO as 1.0 ×10⁹ molecules cm⁻² s⁻¹, kept constant across all simulations. Because we do not increase the flux of NO alongside NO₂, our simulations may be regarded as somewhat conservative. Other fixed boundary conditions of N-bearing species include a flux of 1.53 × 10⁹ for N₂O, a mixing ratio of 0.78 for N₂, and fixed deposition velocities of 2.1 × 10⁻¹ cm s⁻¹ for HO₂NO₂ and HNO₃.

Results from our 1D photochemical model are shown in Figure 1(a). This plot shows the NO₂ volume mixing ratio profiles of an Earth-like planet around the four stellar spectral types we considered in this study: the Sun (blue), AD Leo (green), Proxima Centauri (black), and the K6V star (magenta). Two end-member concentrations are shown: the standard Earth level flux of 8.64 × 10⁹ molecules cm² s⁻¹ (1×, solid curves) and a flux of 172 × 10⁹ molecules cm² s⁻¹ (20×, dashed curves). The corresponding stellar spectra are shown in panel (b), highlighting the wavelength region of strongest NO₂ absorption. As shown in this figure, the hotter stars provide more photons between 0.25 and 0.65 μm, which increases the rate of NO₂ photolysis (Equation (1), and also Table 1).

Zoom In Zoom Out Reset image size

However, photolysis is not the only important factor. As shown in Table 1 and Figure 2, O₃ also plays a major role in determining NO₂ concentration. Figure 2(a) shows the O₃ mixing ratio profiles for various stars. For the Sun (blue), the O₃ concentration increases rapidly below ∼20 km compared to other stars. O₃ participates in the dominant production reaction for NO₂, with the help of NO as shown in Table 1. Ideally, this should increase the concentration of NO₂ below 20 km. However, as shown in Figure 2(b), photolysis of NO₂ due to photons of wavelengths between 0.29 and 0.42 μm that penetrate into the troposphere dominates the destruction of NO₂, decreasing its mixing ratio (see Figure 1(a)). While the photolysis rates generally increase for all stars below 20 km as shown in Figure 2(b), it is the rate at which O₃ increases below this altitude that determines the slope of decrease in NO₂ in the troposphere for planets around different stars. While the rapid increase in O₃ is mostly negated by the rapid photolysis and consequent decrease of NO₂ below ∼20 km for Sun-like stars (blue solid curve in Figures 1 and 2), for other stars the O₃ concentration increases only a little from the surface to the tropopause (black, green, and magenta curves in Figure 2). Consequently, the increasing slope of photolysis rate of NO₂ below ∼20 km for these other stars slightly dominates (panel (b)), resulting in a larger decrease in the mixing ratio of NO₂ compared to a Sun-like star (Figure 1(a)).

Zoom In Zoom Out Reset image size

As a result, the column-integrated NO₂ abundance increases moving from hotter stars to cooler stars (Table 2). The absorption cross sections for NO₂ and other key gases are shown in Figure 3. A Sun-like star produces more photons at wavelengths where NO₂ is photolyzed (between 290 and 420 nm), so the photolysis rate of NO₂ is higher for the planet around the Sun than for a planet around a cooler star (Table 1).

Zoom In Zoom Out Reset image size

The result of all the dominant production and destruction reactions discussed above is that NO and NO₂ decrease with altitude until ∼20 km, into the stratosphere. In the stratosphere, ozone can generate NO₂ via reactions with NO; ozone's overlapping UV cross section with NO₂ also provides some UV shielding. At higher altitudes, above the ozone layer, photochemical processes, including NO₂ photolysis and reaction with OH radicals, draw down abundances. These reactions occur most markedly for the planet orbiting the Sun; NO₂ photolysis proceeds 1–2 orders of magnitude faster around the Sun than around the cooler stars. Reaction of NO₂ with OH to form HNO₃ occurs two orders of magnitude more efficiently around the Sun than around the K6V star, and fully 4–6 orders of magnitude more efficiently around the Sun than around the M-dwarfs.

It is important to note that placing constraints on a planet's NO₂ abundance from its spectrum would not definitively answer whether the NO₂ is biologically or abiotically produced. One would need to estimate the production rates required to produce the observed NO₂ abundance and evaluate whether abiotic sources alone can sustain the inferred production rate.

The absorption cross section of NO₂ shows a broad absorption between 0.25 and 0.6 μm, which has little overlap with absorption from other terrestrial molecular atmospheric constituents (Figure 3(b)). The main possible confusion would be related to aerosols with submicron sizes (∼0.5 μm), which have absorption features that could mimic the exact same shape as NO₂. Considering the broad nature of the NO₂ spectral feature, a unique spectroscopic identification will therefore be ultimately challenging, and this investigation solely explores the hypothetical requirements for a possible detection for an absorption due to NO₂. Other absorption features are also present at ∼3.5 μm, 6.4 μm, and 10–16 μm, but these overlap with absorption bands from H₂O, CO₂, and other species (Figure 3(b)). In order to assess the detectability of NO₂ as a technosignature, we use the mixing ratio profiles from the 1D photochemical model as input to the Planetary Spectrum Generator (PSG,⁹ Villanueva et al. 2018) to simulate reflected light and transit spectra. We estimate the signal-to-noise ratio (S/N) of detecting NO₂ features. PSG is an online radiative transfer suite that integrates the latest radiative transfer methods and spectroscopic parameterizations, and includes a realistic treatment of multiple scattering in layer-by-layer spherical geometry. It can synthesize planetary spectra (atmospheres and surfaces) for a broad range of wavelengths for any given observatory.

We performed simulations with PSG to generate reflected light spectra (Figure 4) of planets around a Sun-like star and a K-dwarf star. We then calculated the S/N required to detect the NO₂ feature (Figure 5) between 0.2 and 0.7 μm. For these simulations, we assumed a LUVOIR-A-like telescope (15 m) observing with the ECLIPS (Extreme Coronagraph for Living Planetary Systems).¹⁰ This instrument is an internal coronagraph with the key goal of direct exoplanet observations. It has three channels: near-UV (NUV, 0.2–0.525 μm), visible (VIS, 0.515–1.030 μm), and near-IR (NIR, 1.0–2.0 μm). The NUV channel is capable of high-contrast imaging only, with an effective spectral resolution of R ∼ 7. The optical channel contains an imaging camera and integral field spectrograph with R = 140. For our spectral simulations, we use NUV (R = 6) and visible (R = 70) channels, as the NO₂ cross section spans from UV into visible wavelengths (see Figure 3). Because the NO₂ feature is quite broad in the NUV to visible region, a low resolution of R = 6 and R = 70 is sufficient to resolve the feature, at the same time maximizing the S/N. We calculated wavelength-dependent S/Ns shown in Figures 4–8 as the difference between the spectra with and without the NO₂ feature, divided by the noise simulated by PSG for the instrument under consideration (see Section 5.3 of Villanueva et al. 2018, and also the PSG website¹¹ where the noise model is discussed in detail). The "net S/N" is calculated by summing the squares of the individual S/Ns at each wavelength within a given band (either NUV or VIS), and then taking the square root. This methodology is largely insensitive to S/N, as long as the feature is resolved by the spectrum. See also the Appendix for a comparison between the PSG coronagraph noise model and a complementary noise model (Robinson et al. 2016; Lustig-Yaeger et al. 2019), showing highly comparable results for the photon count rates and resulting spectral precision. We considered the planets around both the Sun-like and K6V stars to be located at 10 pc distance, residing in the respective habitable zones (HZs) of their host stars as calculated from Kopparapu et al. (2013, 2014), and observed at a phase angle of 45° (0° is secondary eclipse, and 180° is transit). For this feature to be detected, the planet need not be in the HZ, as will be explained later in the discussion (Section 4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In both panels of Figure 4, the difference in the geometric albedo spectrum with and without NO₂ is shown for different levels scaled by factors of current Earth levels in a 10 hr observation with the LUVOIR-15 m telescope. The corresponding noise is shown as a dashed curve. For the Sun-like star (a) even very high concentrations of NO₂ (20× the present Earth levels) barely reach the 1σ noise level in the strongest wavelength region. In panel (b), increasing the nominal abundance to higher concentrations on a planet around a K-dwarf produces only a marginal improvement over a Sun-like star, with the highest concentration (20×) reaching just above the noise level. This is likely because the column number density of NO₂ on a planet around a K-dwarf star seems marginally larger (Table 2) as a result of a lower photolysis rate (last row, Table 1). As discussed above, the enhanced NO₂ absorption on the K-dwarf planet compared to the planet around the Sun-like star is driven by the photochemistry.

In Figure 5, we show the calculated S/N values of the features shown in Figure 4 as a function of wavelength for 10 hr exposure times with a LUVOIR-A-like telescope for wavelengths relevant to the NO₂ feature. The "net S/N" indicated in this figure is calculated by summing up the squares of the S/N from each wavelength band and then taking the square root (see Equation (6) of Lustig-Yaeger et al. 2019). Figure 5(a) shows that for planets around Sun-like stars even an increase of 10× in the NO₂ flux is not enough to detect the feature with any meaningful S/N within 10 hr of observation. Any lower amount of NO₂ would need even longer observation times.

Figure 5(b) shows S/N as a function of the same wavelength range for a planet around K-dwarf star. The combined effects of more NO₂ and better planet–star contrast ratio relative to the planet orbiting the Sun (a K6V dwarf is only about one tenth as luminous as a G2V dwarf) make only a marginal difference in S/N that can be reached in the same time as for a Sun-like star.

While these S/Ns may not look promising, there is an interesting question that one can ask: How much LUVOIR-15 m time is needed to detect the present Earth-level concentration of NO₂ around a Sun-like star at 10 pc? Figure 6(a) shows the spectrum of the difference in geometric albedo with and without NO₂. Also shown as dashed lines are the noise levels for 10 (blue), 300 (red), and 1200 (black) hr of LUVOIR-A observation times. The present Earth level of NO₂ seems to be well above the noise level after 300 hr of observation time (compare the solid green curve with dashed red line), indicating that it might be detectable. To find out with what S/N it would be detectable, Figure 6(b) shows the "net S/N" to detect present Earth-level NO₂ as a function of observation time. To achieve a net S/N of 5 (dashed red line), it would take LUVOIR-15 m about 400 hr. For comparison, to obtain the Hubble Ultra Deep Field (UDF) image, ∼400 hr of actual observation time (∼1 yr in real time) was needed (Beckwith et al. 2006). In fact, Hubble has run even larger programs, such as the CANDLES galaxy evolution survey (Grogin et al. 2011) with 902 orbits (∼900 hr of observation time assuming ∼1 hr per orbit). This took about 3 yr in real time. However, these large programs also obtained data on a huge sample size with thousands of galaxies. LUVOIR is envisaged to be 100% community-competed time, and the final report of the LUVOIR team laid out a Design Reference Mission (DRM) in which comparable allocations of time were spent on general astrophysics observations and exoplanet detection and characterization observations during the first 5 yr of the mission. So, over the course of the nominal LUVOIR mission lifetime of about 5 yr, it may be possible to take data with ∼400 hr observation time on a prime HZ planet candidate(s) within 10 pc, to potentially obtain an S/N ∼ 5 for an NO₂ feature at the present Earth level on an Earth–Sun system at 10 pc. An even more interesting aspect is that we can place upper limits on the amount of NO₂ available on that planet as we spend more observation time on a prime HZ candidate. This could potentially indicate the presence or absence or the level of technological civilization on that planet.

Zoom In Zoom Out Reset image size

While the results from the previous section provide a preliminary study of NO₂ as a potential technosignature, some caveats need to be mentioned. First, we have performed 1D photochemical model calculations using a modern Earth template generated from a 1D radiative–convective, cloud-free climate model from Kopparapu et al. (2013). Clouds can significantly affect the observed spectrum and potentially alter the calculated S/N. To test this, we have prescribed water-ice clouds (particle size 25 μm) between 0.001 and 0.01 bar, and liquid water clouds (particle size 14 μm) between 0.01 and 0.1 bar in PSG. Figure 7 shows S/N as a function of wavelength for an Earth-like planet around a Sun-like star at 10 pc distance observed with the LUVOIR-15 m telescope for 10 hr with (blue solid curve) and without (red solid curve) clouds. The absorption cross sections of water and ice clouds are in the same wavelength region as the peak NO₂ absorption, which further masks the NO₂ feature in this band. We should caution that this is all based on an ad hoc prescription of clouds at a certain height, and a more rigorous analysis using 3D climate models that can simulate self-consistent and time-varying cloud cover needs to be performed. We leave that for future study.

Zoom In Zoom Out Reset image size

Second, we have used the 15 m architecture of LUVOIR-A, and the S/N values we report are a best-case scenario owing to its large mirror size. Other telescope architectures such as LUVOIR-B (8 m) or HabEX¹² (a 4 m mirror accompanied by a coronagraph and a starshade) may need more observation time than shown in Figures 5 and 7 to detect NO₂ features.

As shown in Figure 3, NO₂ also absorbs in the infrared part of the spectrum, particularly in the ranges 3.2–3.7 μm, 5.2–8.9 μm, and 9.7–18 μm. However, the absorption in these regions is either weak across the band compared to the 0.25–0.65 μm visible band, or limited to a very narrow region of the spectrum. Consequently, detecting NO₂ in transit spectroscopy with either JWST or the flagship mission concept study Origins Space Telescope (OST) would be challenging. Nevertheless, we tested this with PSG, and the results are shown in Figure 8. We placed a planet like Proxima Cen b around its host star at 10 pc, assuming that it transits, with 20× NO₂ abundance to maximize the signal. We used JWST NIRSpec and the OST MISC-Transit instrument for the ∼3 μm and ∼6 μm wavelength regions for the detection of NO₂. While OST has greater performance than JWST, the observations here are limited by masking features from H₂O and CO₂ in the near-IR. Even at 20× NO₂ from our photochemical model H₂O features completely dominate the ∼6 μm region of NO₂ absorption (Figure 8(a)). The S/Ns for 10 hr and 500 hr observation times for both telescopes are shown in Figure 8(b). Even with long observation times, it would be very challenging to detect the NO₂ feature with any meaningful S/N.

Zoom In Zoom Out Reset image size

A space-based nulling interferometer such as ESA's LIFE (Large Interferometer for Exoplanets) mission concept (Defrère et al. 2018; Quanz et al. 2018) could potentially detect mid-IR (5–20 μm) features in direct imaging spectra. While we are unable to assess quantitative limits on S/N at this time for this mission, we speculate that phase-dependent thermal emission spectroscopy (Wolf et al. 2019; Suissa et al. 2020) may be another way to detect the NO₂ feature.

Historically, the United States' NO₂ concentrations have varied (gone down) by a factor of 3 over a period of 40 yr, from 1980 to 2019.¹³ Therefore, we can expand the possibilities of detecting a technological civilization at the stage where Earth's civilization was 40 yr ago. It is possible to imagine a more highly industrialized society that could possibly operate in the regime of 5× Earth NO₂ level, making it possible to detect it with LUVOIR-15 m with even less observation time than for present Earth conditions. We should stress here that when we mean a technological civilization, this does not necessarily mean a much more advanced society than the current Earth level. Just like the search for biosignatures encompasses "Earth through time" with different stages in the evolution of Earth's biosphere, we could do a similar search for a "technosphere" at different stages of a technological civilization.

It is possible that atmospheric technosignatures, particularly industrial pollutants such as NO₂, are short-lived. However, this is comparable to searches for radio technosignatures, where the transient nature of the radio communicative civilizations may also be short-lived. Furthermore, it may be that an industrialized society that is prone to emit NO₂ as a byproduct of their combustion technology may also have radio communication capabilities, just like us. In this respect, a search for radio technosignatures can be performed if NO₂ is detected on a potential habitable planet.

If we are looking for NO₂ as a technosignature, and not as a biosignature, then it may appear that one need not limit the search to known planets in the HZ. A technological civilization can possibly inhabit even an adjacent barren planet (like Mars in our solar system), and use the atmosphere as a waste dump of NO₂ emissions. Or they may prefer to live below the surface on a HZ planet and release "waste" NO₂ into the atmosphere. Speculations are endless. However, industrial NO₂ on Earth is essentially produced by burning biomass (coal, petroleum products) that have been excavated to fuel the civilization. (We note that NO₂ can also be produced by nuclear detonations.) The vast majority of burnable organic matter is directly or indirectly derived from oxygenic photosynthesis, meaning that an abiotic or anoxic world would not have abundant preserved organic matter. To burn this biomass, one needs an atmosphere with oxygen. The observation of high abundances of NO₂ on an exoplanet atmosphere would indicate a sustained source of industrial production, likely requiring an oxic atmosphere and indicating a significant source of biomass to sustain long-term industrial activity. While NO₂ can exist in abundant quantities on planets around K-dwarf stars, it may not necessarily be a desirable thing for the inhabitants if they have biology similar to humans, because exposure to NO₂ could cause impairment of lung function and/or recurrent respiratory problems (Faustini et al. 2014). Conversely, if extraterrestrial biology is sufficiently different from Earth life, then it could be impervious to NO₂ toxicity. In this respect, NO₂ on K-dwarfs is similar to the likely accumulation of abiotic and biologically produced CO on Earth-like planets orbiting mid-to-late M-dwarfs, in addition to the accumulation of biosignature gases (Schwieterman et al. 2019).

Missions such as LUVOIR, HabEX, and OST may have biosignature targets as a priority, so it may be untenable to seek dedicated observing time for exclusive technosignature detection. However, in the search for exo-Earth candidates, we will undoubtedly detect other planets within the stellar system (Stark et al. 2014; Kopparapu et al. 2018). LUVOIR and HabEX will be able to simultaneously obtain the spectra of the other bright planets in the system, while performing their observations on a prime HZ target. Consequently, there may not be a need to schedule separate observation time for technosignature detection, because such efforts could "piggyback" on a routine survey to observe both HZ and non-HZ planets (Lingam & Loeb 2019). However, this assumes that NO₂ detection will likely occur within the total integrated observational time spent on the prime HZ candidate for biosignature detection, whereas Figure 5 indicates lower NO₂ abundances may require longer search times.

The presence of NO₂ on Earth today results in part from sustained industrial processes in urban areas. This paper suggests that the detection of NO₂ in an exoplanet atmosphere could serve as a technosignature, because Earth-level biogenic sources would be unable to generate detectable atmospheric abundances of NO₂. Using a 1D photochemical model that uses the present atmospheric temperature profile of Earth, we find that it would be challenging to detect Earth-level NO₂ around G and K-dwarf stars through direct imaging with only 10 hr of observation time. To detect the present Earth-level NO₂ concentration with S/N ∼ 5, it would take ∼400 hr of LUVOIR-15 m telescope. Such large programs may be possible considering several hundred hours of observing time spent on Hubble UDF and CANDLES surveys. Historically, the United States' NO₂ emission has varied (decreased) by roughly a factor of ∼3 over 40 yr from 1980 to 2019. Hence, it might be possible to detect a technological civilization at the stage where Earth's civilization was 40 yr ago with even less time. In this cloud-free model, for habitable planets orbiting K-dwarf stars, by comparison, marginally less time would be needed to detect the present-day NO₂ abundance. The advantage of searching for K-dwarf planets has already been noted in the search for biosignatures (Cuntz & Guinan 2016; Arney 2019), and our results indicate that K-dwarf planets could similarly be advantageous when searching for technosignatures such as NO₂.

However, when we prescribe water-ice and liquid water clouds, there is a moderate decrease in the S/N of the geometric albedo spectrum from LUVOIR-15 m, with present Earth-level NO₂ concentration on an Earth-like planet around a Sun-like star at 10 pc. Clouds and aerosols can reduce the detectability and could mimic the NO₂ feature, posing a challenge to the unique identification of this signature. This highlights the need to perform these calculations with a 3D climate model that can simulate variability of the cloud cover and atmospheric dynamics self-consistently.

While NO₂ absorbs even in the near-IR and mid-IR, we find that it may prove challenging to detect NO₂ in transit observations in this region with JWST and OST because of the weaker absorption and also due to overlapping gas absorption of potent greenhouse gases such as H₂O, CO₂, and CH₄.

Further work is needed to explore the detectability of NO₂ on Earth-like planets around M-dwarfs in direct imaging observations in the near-IR with ground-based 30 m class telescopes. NO₂ concentrations increase on planets around cooler stars due to reduced availability of short-wavelength photons that can photolyze NO₂. Non-detectability at longer observation times could place upper limits on the amount NO₂ present on M-dwarf HZ planets like Proxima Cen b.

The serendipitous detection of NO₂ or any other potential artificial atmospheric spectral signature (CFCs, for example) may become a watershed event in the search for life (biological or technological). Is it likely that biosignatures are more prevalent than technosignatures? We will not know for certain until we search. Our aim in this study is to point out that biosignatures and technosignatures are two sides of the same coin, and searches for both with upcoming observatories can coexist. It is worth pointing out the obvious in this concluding statement: the question "Are we alone?"—which has been the driving force behind the search for extraterrestrial biosignatures—is a question posed by a technological civilization.

The authors would like to thank an anonymous reviewer whose comments greatly improved the manuscript. We would also like to thank Sandra Bastelberger, Chester "Sonny" Harman, Thomas Fauchez, James Kasting, and Avi Mandell for discussions that helped in this work. R.K. would like to acknowledge Vivaswan Kopparapu, his 11 yr old son, who helped R.K. to realize that increasing the LUVOIR-15 m observation time by a factor of 4 doubles the NO₂ S/N for an Earth–Sun system at 10 pc. Goddard affiliates acknowledge support from the GSFC Sellers Exoplanet Environments Collaboration (SEEC), which is supported by NASA's Planetary Science Divisions Research Program. J.H.M. gratefully acknowledges support from the NASA Exobiology program under grant 80NSSC20K0622. This work was performed as part of NASA's Virtual Planetary Laboratory, supported by the National Aeronautics and Space Administration through the NASA Astrobiology Institute under solicitation NNH12ZDA002C and Cooperative Agreement Number NNA13AA93A, and by the NASA Astrobiology Program under grant 80NSSC18K0829 as part of the Nexus for Exoplanet System Science (NExSS) research coordination network.

We conducted a comparison between the LUVOIR coronagraph noise model included in PSG and the Python implementation of the coronagraph noise model of Robinson et al. (2016) from Lustig-Yaeger et al. (2019) (henceforth CG) using Table 3 input values. We found the two models to agree very well (Figure 10), with both models implementing very similar formalisms for computing sensitivities.

We define the end-to-end throughput for the planetary fluxes as T_total = T_Tele × T_cor × T_opt × T_read × T_QE, where T_Tele accounts for light lost due to contamination and inefficiencies in the main collecting area, T_cor is the coronagraphic throughput at this planet–star separation, T_opt is the optical throughput (the transmissivity of all optics), T_QE is the raw quantum efficiency (QE) of the detector, and T_read is the readout efficiencies. The left panel in Figure 9 shows the optical throughput (T_opt) from Stark et al. (2019) and the right panel shows the coronagraph throughput as a function of planet–star separation (T_cor). Although the design of the LUVOIR-A coronagraph has multiple different masks with slightly different Inner Working Angles (IWAs), both coronagraph models use a combined mask (shown in Figure 9) to approximate the optimal use of the coronagraph for any simulated target. Importantly, the coronagraph throughput already accounts for the fraction of the exoplanetary light that falls within the photometric aperture, denoted f_pa in Robinson et al. (2016), so we manually set f_pa = 1 in the CG model to properly account for this factor. The number of stellar photons is defined by the contrast at the core throughput, and thus the number of stellar photons is calculated as $C\times max({T}_{{\rm{cor}}})\approx {10}^{-10}\times 0.27$. For T_Tele, we adopt 0.95 for all wavelengths, on par with the particulate coverage fraction for JWST's mirrors. EMCCD detectors are expected to have T_read near 0.75 (Stark et al. 2019), while for NIR and other detectors, readout inefficiencies and bad pixels may account for a similar value, and we adopt T_read = 0.75 across all detectors as a conservative estimate. The reported QE of the different detectors ranges from 0.6 to 0.9, yet technological improvements in several of these detectors could be expected in the near future, and we adopt a general T_QE = 0.9 for all detectors.

Zoom In Zoom Out Reset image size

The S/N is effectively defined by the different sources of noise, quantified as count rates. We not only compared resulting S/Ns between the two models, but also the simulated count rates for the different components, and found very good agreement (Figure 10). For these simulations, we assumed a circular aperture defined by diffraction (1.22λ/D), an exo-zodiacal level of 4.5 times that of our solar system (22 $\mathrm{mag}\,{\mathrm{arcsec}}^{-2}$), and a local zodi level of 22.5 $\mathrm{mag}\,{\mathrm{arcsec}}^{-2}$. The noise term was computed as

$$\begin{eqnarray}&&{C}_{\mathrm{noise}}=\sqrt{{C}_{p}+{C}_{s}+2{C}_{b}}\end{eqnarray} \tag{ A1 }$$

where C_p is the total number of planet photons, C_s is the stellar photon noise (e.g., "leakage" through the coronagraph), and C_b is the total background, which includes all other noise sources such as zodi, exozodi, dark current, thermal, and read noise. Observations are normally performed as on–off, meaning one with the planet and one without. As such, the background sources of noise need to be counted twice (Equation (A1)). Depending on the observational procedure, the stellar photons can be assumed to be present in the "off" position or not. Robinson et al. (2016) assumes by default that the star is also in the "off" position, and therefore doubles C_s, while the default in PSG is the "off" position without star leakage (so only counted once, Equation (A1)). We explored the impact on the S/N of this assumption in the observational procedure, and only observe small (<10%) differences in the resulting S/N (Figure 10).

Zoom In Zoom Out Reset image size