Introduction

The search for signs of life beyond Earth is an increasingly popular scientific research area as telescope capability advances. For exoplanets, the successful launch and operation of the James Webb Space Telescope (JWST)1, brings in a new level of precision for exoplanet atmosphere measurements, fostering hope that the community may find signs of life through the detection of exoplanet atmosphere biosignature gases within a decade2,3,4,5,6,7,8. Yet, as the community pushes deeper into biosignature gas identification, a dawning conclusion is that biosignature gases will always be stymied by false positives, fatal to robust conclusions about the presence or absence of life. Because of this, there is a growing interest in the study of technosignatures as indicators of life beyond Earth.

Technosignatures are astronomically detectable signs of a technological society on an exoplanet9,10,11,12,13,14. The concept of technosignatures dates back decades and includes several categories, such as: Lowell’s canals on Mars15; artificial illumination of a planet (e.g.,16,17; megastructures (e.g.,18,19); waste heat (e.g.,18,20,21); and artificial (i.e., non-terrestrial) artifacts (e.g.,22,23,24). Technosignature research has recently been growing, e.g., 11,12,25,26.

Technosignature gases, as a subset of technosignatures in general, are artificially produced gases that can accumulate to detectable levels in an exoplanet atmosphere. A technosignature gas can either be emitted for a specific purpose or released as a by-product of industrial civilization12,16,27. Proposed technosignature gases center around industrial pollutants such as chlorofluorocarbons (CFCs), hydrofluorocarbons (HFCs), and perfluorochemicals (PFCs)12,28. A conclusion from the above references is that detection of CFCs in exoplanet atmospheres, under the most favorable assumptions, would require 100–500 h of JWST in-transit observation time (Table 1). This is increased to 200–1000 h of total observing time, when taking into account an out-of-transit equivalent baseline time. This can be compared to the current JWST time allocation to individual exoplanets which are typically up to 20 h, with a few planet full phase curves allocated around 40–50 h, and one transiting exoplanet atmosphere exception at 70 h29. The exception to the long-required observation times is for the hypothetical and highly favorable case of a terrestrial-size planet transiting a tiny star—a white dwarf star which itself is about the size of Earth and is a dead remnant of a Sun-like star. This scenario would require only tens of hours of in-transit time28,30,31. See the SI for more details on CFCs and other suggested technosignature gases.

Table 1 Summary of proposed exoplanet atmosphere technosignature gases including simulated in-transit observation times.
Full size table

Here we propose nitrogen trifluoride (NF3) and sulfur hexafluoride (SF6)—fully fluorinated non-carbon compounds—as potential technosignature gases (Fig. 1). A fully fluorinated compound is a molecule where the central atoms are only bonded to fluorine atoms. For example, in NF3 the central nitrogen atom can bind to three other atoms. Because each of the bound atoms is fluorine, we call NF3 fully fluorinated. NF3 and SF6 have been only briefly mentioned as technosignatures34,35, the case has not previously been developed.

Figure 1
figure 1

Chemical schematic of the fully fluorinated molecules: SF6 (left) and NF3 (right). Atoms are depicted by colors as follows. Yellow: sulfur. Blue: nitrogen. Pink: fluorine.

Full size image

We are motivated to explore non-carbon fully fluorinated compounds primarily because life on Earth never makes compounds containing the N–F and S–F bonds, nor fully fluorinated non-carbon compounds ("Results" section). S–F and N–F bonds are not even made as metabolic intermediates and are likely to be universally excluded by life, no matter its biochemistry. In fact, life on Earth very rarely uses F chemistry. Only a few species make C–F bonds at all (Petkowski et al. in prep.). While life on Earth does not make fully fluorinated carbon molecules (e.g., tetrafluoromethane, CF4, or CFCs that have been previously considered as potential technosignature gases12), life could in principle do so without inventing completely new enzymes and other necessary biochemical repertoire (Petkowski et al. in prep.). Hence fully fluorinated non-carbon molecules may be more robust technosignature gases than (fully) fluorinated carbon molecules.

SF6 and NF3 on Earth are not only industrial pollutants but their relative atmospheric abundance has rapidly increased36,37 (See Fig. 2 and SI). NF3 has no known abiotic sources and was entirely absent from the pre-industrial atmosphere37. Within the last half century, the NF3 abundance in the atmosphere rose to close to 3 part-per-trillion (ppt) by volume (Fig. 2). SF6 does not have significant abiotic sources that could mimic its rapid increase in the atmosphere. Like NF3, the steady and rapid relative increase of SF6 in the atmosphere from very low background abundances of < 0.06 ppt to around 11 ppt in the last half a century combined with its relatively long atmospheric lifetime of at least a couple of hundred years38 further supports SF6 as a good technosignature gas candidate. However, we caution that the atmospheric chemistry of NF3 and SF6 has not been well studied and many potential destruction pathways for NF3 and SF6 may not be known (see Tables S2 and S3).

Figure 2
figure 2

Atmosphere gas abundance for some industrial pollutants including NF3 and SF6. The y-axis is the fractional gas abundance in part-per-trillion (ppt) by volume and the x axis is time in years. Both NF3 and SF6 have a rapid increase compared to other industrial pollutants. The right panel is a zoomed in version of the left panel. Data from36,37,95, and notably 2013–2022 data is from the Global Monitoring Laboratory (GML) (SF6: https://gml.noaa.gov/hats/combined/SF6.html; NF3: https://gml.noaa.gov/hats/gases/NF3.html).

Full size image

Results

No fully fluorinated molecules are made by life

Life on Earth is not known to produce any molecules with N–F or S–F bonds, and this includes fully fluorinated N and S compounds. We derived this result from our natural products database which is a curation of all known biochemicals and natural products (i.e. compounds produced by life) from an extensive literature online chemical repository search ("Methods" section and39,40). Here, "natural products" means chemicals made by life.

Life does produce some compounds with N–Cl, N–Br, S–Cl, and S–Br bonds, but none are volatile. In addition, the N–Cl, N–Br, S–Cl, and S–Br compounds are typically intermediaries and not molecules that accumulate on their own. The molecules containing N–Br and N–Cl are quite reactive and therefore rare, a notable example being a neurotransmitter N-bromotaurine41 and pseudoceratonic acid42. The S–Cl and S–Br bond-containing molecules have been found only in proteins as intermediaries in synthesizing the N–S bonds39.

In comparison with the rare N–Cl, N–Br, S–Cl, and S–Br containing molecules, there are thousands of compounds containing C–Cl, C–Br and C–I bonds that are made by life (Fig. 3). Most of the life-produced halogenated compounds are Cl-containing compounds (~ 2% of all known natural products, where the current known total number of unique natural products is ~ 220,000). Br-containing compounds produced by life are a close second (~ 1.7% of all known natural products). Iodine-containing natural products are much more rare but still a significant group with approximately 200 known examples (~ 0.1% of all known natural products). The above includes solids, liquids, and gases; it is worth noting that all of the volatile halogenated (Cl-, Br, I-containing) compounds produced by life are halocarbons, where the halogen atom is directly bonded to carbon.

Figure 3
figure 3

Number of molecules containing C–X bonds produced by life (called “natural products”), where X is Cl, Br, I, or F. For comparison the numbers are separated into three categories. Green: all known natural products in the category (i.e. produced by life). Blue: the subset of volatile natural products, here limited to molecules with 6 non-H atoms or less. Red: the subset of fully halogenated volatile natural products, also limited to molecules with 6 non-H atoms or less. Life rarely produces fully non-F halogenated volatile natural products and does not produce any fully fluorinated products of any kind (C–F, N–F, S–F, or other). Note that one fully halogenated molecule is double counted as it contains both Cl and Br, bromotrichloromethane. Not shown is that life does not produce any molecules with N–F or S–F bonds.

Full size image

In contrast to the thousands of Cl-, Br- and I-containing carbon compounds made by life, the F-containing compounds are nearly excluded from life’s repertoire, numbering only 34. And, only two of these 34 are volatile. Nearly all of the known biogenic F-containing natural products are fluorinated carboxylic acids (Petkowski et al. 2023 in prep.). There are no fully fluorinated C–F compounds known to be produced by life.

Although life does not produce fully fluorinated molecules, life does actually produce at least four carbon compounds that are fully halogenated with halogens other than fluorine. This is a very small percentage of all volatile halocarbons produced by life. For some numbers, and considering molecules with 6 or fewer non-hydrogen atoms, there are 85 volatile halocarbons produced by life. There are 34 possible halomethanes, of which 14 are fully halogenated. 22 halomethanes are produced by life on Earth40. Out of those 22, only 3 are fully halogenated: tetrachloromethane (CCl4) produced by several plants and marine algae43, tetrabromomethane, CBr4, produced by from various marine algae, e.g. Asparagopsis taxiformis44, and bromotrichloromethane, CBrCl3, that contains both Cl and Br atoms, produced by marine algae44. For completeness, the fourth fully halogenated carbon compound produced by life is tetrachloroethene (C2Cl4), produced by Hawaiian red seaweed Asparagopsis taxiformis44.

Again, the number of fully fluorinated molecules made by life is zero, no matter if they are fluorocarbons or if the F atom is attached to a non-carbon element.

We explain why life avoids F-containing compounds in Petkowski et al. 2023 (in prep.) and briefly summarize the explanation here. We have identified three challenges that F chemistry poses for life on Earth that make the use of fluorine in Earth’s biochemistry a difficult prospect:

  1. 1.

    Relatively low bioavailability of F, which is primarily locked inside insoluble minerals, and is not available in surface water (unlike Cl, Br and I).

  2. 2.

    The uniquely high electronegativity of F, which means that enzymes that handle Cl, Br, peroxide and other oxidizing species cannot be repurposed to handle F. As a result of this challenge life needs to evolve a completely novel enzymatic machinery to create C–F bonds.

  3. 3.

    The lack of reactivity of the C–F bond, which makes evolving catalysts that can handle C–F bonds a difficult task.

These substantial evolutionary barriers mean that almost all life has found ways to address its ecological requirements with chemistry other than C–F chemistry. All three factors are general properties of fluorine chemistry, are not specific to terrestrial biochemistry and therefore are likely universal.

NF3 and SF6 have unique spectral features compared to dominant atmospheric gases

The gases NF3 and SF6 have unique spectral features as compared to bulk terrestrial planet atmosphere gases (Figs. 4 and 5). NF3 and SF6 absorption fall in the 9–12 micron spectral window where the plausible dominant super Earth or Earth-sized planet atmospheric gases CO2, CO, CH4, and the strong H2O vapor spectral features do not appear. Recall the gases N2 and H2 have no distinctive spectral features at infrared wavelengths; being homonuclear they have no net dipole moment. We note that the NF3 and SF6 strong absorbing power in a unique part of the spectrum compared to other major atmospheric gases is why they are potent greenhouse gases here on Earth. Trace gases of interest (such as PH3, NH3, SO2, H2S) other than the dominant terrestrial planet gases also do not have overlapping spectral features to NF3 and SF6 (Fig. 4).

Figure 4
figure 4

The molecular spectra phalanx plot compares the absorbance spectra amongst molecules. The x axis is wavenumber ranging from 4000 to 500 cm−1 (2.5–20 μm). The y axis is the order of the molecules. Color represents the intensity of the absorption peaks; yellow and green represent strong absorption, while blue and purple represent weak or no absorption. Note the absorbance is normalized to 1. The spectra phalanx provides a visualization of which molecules are clustered together in wavelength based on their spectral data. Functional groups are labeled at their clustering points. The grey dashed lines near the top of the plot mark NF3 and SF6 and show their relative uniqueness in wavenumber space as compared to other halogenated species. The spectra data are from 600 gas species in our All Small Molecules Database (ASM)40 (volatile molecules with up to 6 non-H atoms) that have available spectra in NIST64. See “Methods” for details on construction of this figure.

Full size image
Figure 5
figure 5

Simulated spectra of an exoplanet with transiting an M5V star with an atmosphere abundance of 1 part-per-million of NF3 and SF6. Top panel: the y-axis shows transit depth (ppm), and the x-axis shows wavelength (μm). The spectra are simulated from 1 to 23 μm, covering the wavelength span of most of JWST’s observation modes. The yellow, green, and blue regions show the spectral coverage of NIRSpec, and the red region shows coverage of MIRI LRS. Bottom panel: absorption cross sections in cm2 as a function of wavelength for key molecules of interest. For context, the contemporary Earth abundance levels are 3 ppt37 and 11 ppt38 for NF3 and SF6 respectively.

Full size image

There are gases with overlapping spectral features to NF3 and SF6, and these are primarily halogenated carbon compounds (Fig. 4). The multitude of possible halogenated carbon compound gases (which could be biosignatures or technosignatures) pose a more complicated situation. Disentangling SF6 and NF3 from the multitude of possible chlorofluorocarbon gases and those gases from each other, depends on spectral resolution. For SF6 there are only a few candidates with similar main spectral peaks at the same wavelengths to NF3 and SF6. More work is needed to ascertain what is needed to distinguish amongst all of the halogenated gases, considering spectral features of other molecules.

Atmospheric concentrations needed for detection

On the order of 1 part-per-million (ppm) atmospheric abundance by volume of NF3 and SF6 produces a ~ 30 ppm signal in simulated transmission spectra for a terrestrial planet transiting an M dwarf star (Fig. 5). This is for an H2-dominated atmosphere. For a CO2- or N2-dominated atmosphere, the signal is much lower, owing to the much smaller scale height for high mean molecular weight gases compared to the low mean molecular weight gas H2 (Fig. 5). This finding is consistent with other biosignature gas detection simulations, including the point that detecting a ~ 30 ppm signal will almost certainly take tens of transits or more (e.g.,2,45,46,47,48), a much larger number than the typically few transits currently allocated for exoplanets with the JWST29. For details of simulated detectability including JWST noise floor see, for example4,49,50.

However, we again emphasize a major point in favor of NF3 and SF6 spectral distinguishability in an exoplanet atmosphere is that the 9–12 micron region has no expected major atmosphere gases with strong spectral features—although many trace gases, especially halogenated compounds, have features in this window.

The atmospheric accumulation of NF3 and SF6 is favorable due to their low water-solubility (for a comparison to other gases such as CO2 and NH3 see Fig. 6). The low water solubility means they will not dissolve in rainwater and fall to the ground or the sea. The NF3 and SF6 and long atmospheric lifetimes also favor their accumulation (for destruction rates see the SI).

Figure 6
figure 6

Solubility of various atmospheric gases in water. The x-axis shows the chemical species’ names, and the y-axis shows the Henryʼs law constant on a log scale. Henry’s law is defined as HCP(X) = C(X)/p, where HCP(X) is Henry’s law constant for a species X in mol Pa−1 m−3. P is the partial pressure of that species in Pascal, and C(x) is the dissolved concentration (in mol m−3) under the equilibrium condition. The larger HCP, the more soluble the species is. NF3 and SF6 have very low water solubility.

Full size image

Proper estimation of the lifetime of NF3 and SF6 would require experimentally measured kinetics of chemical reactions of NF3 and SF6 in H2 at various temperatures. The current thermochemical literature and databases (such as NIST) have very scarce information on reactivity of those species with relevant atmospheric components, at relevant temperatures (Table S1 and S2). Measurement or detailed modeling of those values is essential for progress.

Nonetheless, we can make some estimates. Regarding the lifetime of NF3 in an H2-dominated atmosphere, the rate constant [cm3 molecule−1 s−1] for the reaction H + NF3 → NF2 + HF at 300 K has been calculated to be around 2.4 × 10−2051, which is four orders of magnitude lower than the rate constant of the reaction with the hydroxyl radical (OH), 4.0 × 10−16, (reaction OH + NF3 → F + H2O + NO2 in Table S1). This difference suggests that the reaction with OH would dominate over the reaction with an H radical as a main destruction pathway for NF3. As a result, the likely lifetime of NF3 in an H2-dominated atmosphere would not be significantly different than it is in the oxidized, OH-rich, atmosphere of Earth. Both Earth’s atmosphere and the reduced H2-dominated atmosphere are expected to be abundant with OH radicals due to photolysis of H2O.

Regarding the lifetime of SF6, it is a stable and unreactive gas with a lifetime of hundreds to thousands of years in Earth’s oxidizing atmosphere. SF6 likely has a similarly long, if not longer lifetime in a H2-dominated exoplanet atmosphere. Experimental shock tube dissociation studies of SF6 in the presence of H2 gas in the temperature range of 1734–1848 K supports this conclusion. The measurements show that H2 does not increase the dissociation of SF6 as compared to argon gas control52. This result indicates that SF6 should have at least a similar lifetime in the H2-dominated atmosphere as it has in Earth’s atmosphere.

Abiotic sources and a false positive mitigation strategy

NF3 has no known abiotic sources

NF3 has no known abiotic sources. In other words, NF3 is not known to be a product of any photochemical, volcanic, or other geological process. We further show formation of NF3 is thermodynamically unfavorable for terrestrial planet conditions (Fig. 7). NF3 is also not known to be released from any fluorine-containing minerals. The absence of false-positives for NF3 is supported by the lack of detection of this gas in any available pre-industrial samples37. With no know abiotic or biotic sources, NF3 is a, if not the, prime candidate for a technosignature gas search.

Figure 7
figure 7

The free energy of formation of NF3, SF6, and SiF4. The y-axis is the standard free energy of formation and the x-axis is temperature (K). The unfavorable thermodynamic conditions (positive ∆G) of the formation of SF6 and NF3 at all relevant temperatures are in contrast to the favorable abiotic formation of SiF4 (negative ∆G), where only SiF4 is a known volcanic product on Earth. The formation of NF3 and SF6 is highly thermodynamically unfavorable and therefore unlikely to be anything but a trace product of planetary geology or volcanism.

Full size image

Abiotic sources of SF6

On Earth trace amounts of SF6 exists in volcanic rocks in rift zones, faults, igneous intrusions, geo-thermic areas and diagenetic fluids53. SF6 is predominantly present in fluorites and some granites, while, for example basalts do not contain detectable SF653,54.

The exact mechanism of abiotic formation of SF6 on Earth is unknown. It is also unclear if SF6 is directly made by volcanoes on Earth or its release is just associated with volcanic activity. Harnisch and Eisenhauer have examined the gases from several volcanic fumaroles, e.g., Etna (Sicily), Vulcano Island (Sicily), Kuju (Japan), and Satsuma Iwojima (Japan), and find that they are not significant sources of SF654. However, the authors note that the underlying rocks of these volcanoes are not granitic and as a result might lack a source for SF654. The equilibrium pre-industrial atmospheric concentration of SF6 on Earth is estimated to be < 0.06 ppt53. The dominant F-containing volcanic gas on Earth is HF with abundances reaching 0.5–15 ppb55. Other trace F-containing species, including NH4F, SiF4, (NH4)2SiF6, NaSiF6, K2SiF6, KBF4, and organo-fluorides, are also associated with volcanic activity or released by volcanoes, but to a much lower extent than HF56,57,58,59. It is therefore unlikely that SF6 will be a significant false-positive on a water-rich terrestrial planet as Earth.

The hypothesis that volcanic SF6 is negligible on Earth is also supported by thermodynamics of SF6’s formation (Fig. 7). The formation of SF6 (and NF3) is highly thermodynamically unfavorable and therefore unlikely to be a source of anything but a trace product of planetary geology or volcanism. The unfavorable thermodynamic conditions of the formation of SF6 (and NF3) are in contrast to the abiotic formation of SiF4, another non-carbon fully fluorinated gas, that is a known volcanic product on Earth60 (Fig. 7).

Strategies to rule out SF6 (and NF3) false positives

We first ask if SF6 could be a significant abiotic gas on a planet with an environment different from Earth. The answer is yes, as follows. On a dry planet, or a planet that is otherwise H-depleted one would expect a different profile of F-containing volatiles erupted by volcanoes than on Earth which could lead to a potential false-positive interpretation of the source of the detected SF6. On a H-depleted planet, fluorine will be bound to a greater extent to elements other than H. Therefore HF, while still expected to be erupted by volcanoes, would not dominate volcanic gases. Such a scenario opens the possibility for SF6 to be a much more abundant volcanic product on dry exoplanets than it is on wet planet Earth.

The view that SF6 could be a volcanic product on an H-depleted world is supported by the tentative detection of 0.2 ± 0.1 ppm SF6 in the atmosphere of Venus by Venera 1461. If correct, this value is five orders of magnitude larger than the amount of SF6 detected in Earth's atmosphere. Since Venus’ crust and atmosphere are significantly H-depleted (though the deeper mantle may be relatively less H-depleted62), it is likely that the majority of F is erupted as other compounds than HF, e.g. SSF2, COF2, FClCO, and SOF2, etc.63. Therefore, it is not unexpected that in the H-depleted environment of Venus, with abundant sulfur, SF6 could also be a volcanic product released in significantly higher abundance than on Earth. SF6 could also be the result of weathering of fluorite minerals which abundance on Venus is poorly constrained.

We now turn back to non-H-depleted planets that are the focus of this paper, and propose a strategy to rule out any volcanic origin of SF6. We propose simultaneous observations of SiF4 to be employed as a method to rule out volcanic sources of SF6 (and NF3). The overview reason is that SiF4 is much more thermodynamically favored over SF6 (see SI Section 4) such that any volcanic activity that produces SF6 will produce significantly higher amounts of SiF4. In more detail, SiF4 is a known volcanic gas on Earth. Volcanic production of SiF4 can reach several tons per day and in some instances, such as in the Satsuma-Iwojima volcano plume, SiF4 production can rival that of HF60. As with SF6, the formation of SiF4 will be favored in hydrogen-depleted environments and low temperature gas sources. It is conceivable that on a planet with much more active low-temperature volcanism and lower crustal and atmospheric H abundance than Earth that a larger quantity of SiF4 would be released by volcanoes into the atmosphere. SiF4, therefore, can be an indicator of geological activity on a planet with volcanic chemistry much more favorable to forming SF6 than Earth.

In an event where SF6 is detected but simultaneous observations of SiF4 is not, the likelihood increases that SF6 is biological or technological. This conclusion is supported by our calculations that suggest that there are no plausible conditions (pressure 0.1–10,000 bar, H2O content 0.1–95%, mantle redox state (MH vs. IW etc.)) where the amount of SF6 (or NF3) produced by volcanoes comes to within ten orders of magnitude of that of SiF4, effectively ruling out the possibility of volcanically-driven co-existence of SiF4 and SF6 (or NF3) (Figure S1 and Figure S2). All of the above rationale applies to NF3; unlike SF6 there are no abiotic sources on Earth, and so volcanism is even less likely to make this gas on another planet.

There appears to be no spectral information spectral information for SiF464, getting such information is crucial for the execution of our proposed mitigation strategy. We conclude this section with a call to study the spectroscopy of the fully fluorinated non-carbon molecules.

Discussion and summary

Prospects

The JWST has been operational for science as of 2022 and is our best currently existing capability for exoplanet atmosphere observations for transiting planets via transit transmission spectroscopy. Two categories of terrestrial planet atmospheres are accessible by JWST. The first is the not yet existing terrestrial planets transiting white dwarf stars (See "Introduction" section). The second category is terrestrial planets transiting small red dwarf stars. In addition to the long-required observation times for technosignature gases for such systems ("Introduction" section), a major challenge is the red dwarf host star variability (e.g.65 and references therein). Stellar variability is a collective term for a rich set of physical phenomena caused by stellar surface inhomogeneities which come in the form of granulation and magnetic features such as spots and faculae (e.g., 66). The host star variability changes with time as active regions evolve and as the star rotates. Stellar spots and faculae have different temperatures from the disk-averaged photosphere, and for cooler stars, have molecular features distinct from the star itself but similar to those in a planet’s atmosphere (65 and references therein). Stellar surface inhomogeneities can induce rogue features in stellar data which mimic signatures of exoplanet atmospheres; some studies have found that the signal from stellar inhomogeneities exceeds the signal from the planetary spectral features (e.g.,65,67,68). A community of over 100 experts has summarized the challenges and potential mitigation strategies of M dwarf host star variability effects on small transiting exoplanet atmospheres (NASA's Exoplanet Exploration Program Study Analysis Group 21 (SAG21)65).

New large ground-based telescopes now under construction are planned to be online within the next decade: the Extremely Large Telescope (EELT, 39 m aperture diameter)69,70; the Thirty Meter Telescope (TMT, 30 m aperture diameter)71; and the Giant Magellan Telescope (GMT, 20 m aperture diameter)72,73. These large telescopes can directly image habitable-zone planets orbiting (i.e., not necessarily transiting) M dwarf stars with the right coronagraph instrumentation and extreme adaptive optics. The challenge is overcoming the high planet-star contrast at 107–108 levels. METIS74 on the EELT will have mid-infrared direct imaging capability via a coronagraph and extreme adaptive optics that includes low and medium resolution spectroscopy; however METIS is designed for planet discovery via imaging and its limited sensitivity means METIS is unlikely to enable reliable ~ 10 micron spectrum of a temperate, rocky world orbiting an M dwarf star (74 and Quanz, priv. comm. 2023). Direct imaging in the thermal infrared for about five Sun-like stars may also be possible74,75. Near infrared imaging and spectroscopy of reflected light from rocky planets in their host star's habitable zone is anticipated to be possible for up to 100 nearby low-mass stars (primarily mid M dwarf stars)75,76,77; while the near-IR is outside of the wavelength range of interest for NF3 and SF6, such programs may help discover new suitable planets for follow up atmosphere observations. Other than direct imaging, the large ground-based telescopes might be capable of a combination of high-dispersion, high-spectral resolution (R ~ 100,000) spectroscopy with moderate high-contrast imaging to observe spectra of a few rocky planets orbiting Sun-like stars78. (Note that not all IR wavelength regions are not easily accessible from Earth’s surface due to Earth’s atmospheric gases).

NASA’s Habitable Worlds Observatory (HWO; https://www.greatobservatories.org/irouv) is a NASA Great Observatory intended to be ready for launch in the mid-2030s and will be designed to directly image exoplanets orbiting Sun-like stars for both discovery and atmospheric characterization. This telescope will have a primary mirror about 6-m in diameter and is planned to have a coronagraph to block out starlight so the planet can be directly imaged. Because HWO will operate at visible to near-IR wavelengths, HWO will not be able to access the infrared windows (> 4 microns) appropriate for NF3 and SF6 spectral features.

The space-based interferometer under study called the Large Interferometer for Exoplanets (LIFE;79) is designed to observe at 4–18.5 microns, and this includes the infrared windows for NF3 and SF6 spectral features (Fig. 5). LIFE would have four elements each with aperture 2–3.5-m diameter as well as a combiner spacecraft80.

Are SF6 and NF3 alien biosignatures or universal technosignature gases?

The exclusion of N–F and S–F compounds from Earth biochemistry, the low water solubility of NF3 and SF6, their unique spectra, long atmospheric life time, and industrial utility on Earth all makes NF3 and SF6 attractive targets as technosignature gases.

We discuss the reasons for and against NF3 and SF6 being produced by an alien biology versus by a technological society.

Life on Earth has apparently never made an N–F or an S–F bond-containing molecule, despite life independently evolving F-handling enzymes several times over the last three billion years (Petkowski et al. 2023, in prep.). Life also rarely makes fully halogenated molecules, with only four known examples ("No fully fluorinated molecules are made by life" section). This suggests that there is at least one major evolutionary barrier to making NF3 or SF6, which would require a correspondingly powerful selective benefit to overcome.

SF6 is a completely chemically inert gas, making it essentially “invisible” to biochemistry. Therefore, we have to ask why life, no matter its biochemical makeup, would invest significant metabolic machinery and energy into making a compound that it then throws away, and which, due to its chemical inertness, has a negligible effect on the organism’s immediate environment or on its competitors. Biological production of SF6 would cost an organism a large amount of energy while not giving any evolutionary advantage (i.e., does not provide any obvious fitness gain).

NF3 is mildly reactive, and so we may speculate that NF3 could have the same ecological role as CH3Br on Earth, as a mildly reactive, non-specific, diffusible toxin to intoxicate the competition in the immediate environment. Therefore, NF3 production could increase the fitness of the NF3-producing organism. The toxicity of NF3 can be to an extent similar to the toxicity of phosphine (PH3)81. PH3 does not react all that much with biological tissues, but it is toxic to humans, and many other O2-dependent organisms, because it converts hemoglobin to methemoglobin (which cannot bind oxygen), i.e. PH3 is toxic to (large, terrestrial) oxygen-dependent organisms but not to anaerobic ones82. Oddly, in identifying a technosignature gas, a biological source may be a false positive. Against this, we note that there are many mildly reactive, toxic carbon-containing compounds that can be made with the enzymatic machinery that any carbon-based life is likely to possess, such as CH3Br, CH3I, cyanogen, CO, formaldehyde, nitric oxide, so NF3 would not provide any unique advantage over these more evolutionarily accessible substances.

One could argue that the biological repertoire of gases on other planets may surprise us because we have no idea what gases non-Earth-like life might produce40. Production of certain biochemicals is often an evolutionary accident or depends on the planetary environmental history. An example is the gas stibine, SbH3, which would not be expected to be made by Earth life because Sb itself is a rare element in the Earth’s crust, but nevertheless is synthesized by terrestrial anaerobic sewage sludge microflora83. Another example is trimethylbismuth (C3H9Bi) produced by a variety of bacteria in anaerobic conditions (reviewed in84). But we favor the point that while it is possible that an alien biochemistry would find use for NF3 and SF6, virtually any combination of physical and chemical properties of NF3 or SF6 can be duplicated with less energy and less dangerous radicals using the chemistry of other halogens and carbon. Indeed, our premise of promoting NF3 and SF6 as technosignatures is that F is nearly excluded by life on Earth (Petkowski et al. 2023, in prep.)) and that such an exclusion may well be universal.

We further argue that industrial use of some chemicals may be likely to be universal. SF6 in particular has unique properties that make it useful for a technological civilization (reviewed "Results" section) and specifically its high dielectric constant and high breakdown voltage as a gas make it valuable in high voltage electrical equipment36. Biosignature gases are the product of evolutionary contingency and are limited by thermodynamics and the reactivity of materials to water. Industrial chemicals, however, are the result of an informed, systematic search of all possible chemicals for materials that have optimal properties for a specific application, largely regardless of the thermodynamics of their synthesis or whether their synthesis requires chemistry that is incompatible with earth surface conditions (for example, industrial production of NF3 is done in molten ammonium fluoride85, an obscure material unlikely to occur on its own on any rocky planet). It is therefore plausible to suggest that an extraterrestrial civilization that wanted, for example, a gaseous product to act as a high voltage insulator and arc quencher, would choose SF6, no matter what the entity's own biochemistry, evolutionary history or planetary environment was.

If we in the future have a way to detect the tiny part-per-trillion amounts elsewhere that we humans have accumulated in our atmosphere, and are lucky to catch the small window where a society becomes industrial, we may observe the rapid, steady increase of the atmospheric abundance of SF6 or NF3. A rapid, steady increase would favor a technological source and may be a solid discriminator between bio- and a true technosignature gas. Even the relatively rapid increase in atmospheric O2 that led to the Great Oxygenation Event still took 1–10 million years for O2 to accumulate in high enough concentrations to have a weathering effect on rocks (e.g.,86). A multi-generational observational campaign can further distinguish between SF6 and/or NF3 as technosignature and not biosignature gases. Observation of a planet over the time span of 2–4 generations (~ 100 years) to monitor the increase of the SF6 and NF3 gases could strengthen attribution to a technological source, but one would have to get lucky with timing.

We now turn to the requirement that the atmospheric abundance of NF3 and SF6 needed for detection with the JWST is of the order of 1 ppm, far higher than current atmospheric levels. Here on Earth, no uniquely technological gas has been produced to accumulate to such relatively high amounts in the atmosphere. 1 ppm SF6 is 100 times the current terrestrial level, and it would take ~ 700 years of the current emission rate to build that concentration in Earth's atmosphere, taking the SF6 atmospheric lifetime of 3200 years. The greenhouse gas effect of such concentrations on Earth would be catastrophic, which suggests that any alien technological society would curtail their emissions before they reached 1 ppm, unless their goal was substantial global greenhouse heating. Indeed SF6 has been considered as a terraforming agent on Mars, albeit for humans at our stage of development a prohibitively expensive one87.

Fully fluorinated non-carbon containing gases other than NF3 and SF6 have little known about them (see discussion in the SI; Table S3).

Summary

In summary, NF3 and SF6 are appealing technosignature gases primarily because use of F is nearly excluded by the chemistry of life on Earth for fundamental reasons (Petkowski et al. 2023, in prep.), and moreover because molecules with N–F and S–F bonds are not made by life on Earth at all. We have argued that while an alien biochemistry might find use for NF3 and SF6, such use would have to provide a significant evolutionary gain that offsets e.g. the large energy expenditure for the synthesis and breakage of fluorine-containing bonds. In contrast, industrial use may be universal due to unique physical and chemical properties of NF3 and SF6 gases (see SI). Therefore, NF3 and SF6 may be likely to be used by alien industry no matter the biochemical makeup of the alien biology or the particular environmental conditions of the alien planet. We note however that it is likely that the technological stage of the civilization, or time-span at which NF3 and SF6 are produced in sufficient amounts to be observed is short.

NF3 has no known sources other than industrial, while SF6 is produced abiotically in extremely small, trace amounts. Their source can be distinguished from e.g. potential volcanic release into the atmosphere by comparison with the known volcanic gas SiF4. The lack of known false-positives for NF3 and very low abundance of abiotic sources of SF6 on Earth supports their potential as a technosignature gases on exoplanets. NF3 and SF6’s long atmospheric lifetimes and unique spectral features aid their potential detection.

Methods

Custom natural products database

To understand the uniqueness of fully fluorinated compounds out of all chemicals made by life (called “natural products”), we use our database of natural product chemicals curated over the last decade. Our database is presented in40 and expanded and completed as described in39. We created and curated our database by an extensive literature search and by searching available online natural product repositories39.

Our natural products database has been rigorously screened to contain only compounds that are a result of natural biochemical processes of a living organism. It also contains biological sources identified for every molecule (i.e., a list of species from which the natural product was isolated).

We emphasize how challenging it is to compile a complete list of all that is known about each natural product. First, no individual database covers more than 20% of the known natural products. Second, because most natural product databases focus on drug design, they include synthetic derivatives of natural products or drug metabolites that often mimic natural molecules while not being true natural compounds themselves. Other databases often include artificial compounds that emerge as a result of “feeding” an organism with a precursor molecule, or completely artificial compounds that have accumulated as contaminants in plants and animals. We manually excluded such compounds from our database. Third, most data sources needed extensive checking and modification due to a range of format differences and coding errors. All of the above problems motivated us to curate our own natural products database (see39 for a full description).

Spectra visualization tool

A key question for any atmospheric trace gas is whether or not its spectral features are distinguishable from features found in other expected atmospheric molecular gases. We introduce a new tool for analysis, detailed in88 and summarized here.

We call our new tool a spectra phalanx plot (Fig. 4). This plot enables a visual comparison amongst the absorption peaks of each individual molecule’s spectral features, with molecules that share similar spectral features grouped closer together. Each molecule occupies a horizontal line parallel to the x-axis, where the molecule’s spectral features are plotted as a function of wavelength location. We use color to represent the intensity of the absorption peaks; yellow and green represent strong peaks, while blue and purple represent weak or no absorption. Note the absorbance is normalized to 1, so only the ratio between the peaks and the strongest peak is important, and comparing the absolute intensity between molecules is not possible. For this tool, we use the ~ 600 transmission/absorbance data spectra available from NIST64 (with a small subset from HITRAN89). The order of the molecules is not meaningful other than that molecules with similar wavelengths of spectral features are grouped together.

We generate the molecule ordering in the spectra phalanx plot using hierarchical clustering, a tree-based approach that builds clusters by iteratively grouping two of the closest cluster/elements into the same cluster and organized in a binary tree structure called a dendrogram90. We apply hierarchical clustering on molecular spectra to cluster the molecules and validate that the molecules in the same clusters share similar chemical structures using a molecular maximum common substructure search using methods described in88.

In more detail, comparing the molecular IR spectra of two molecules expresses how similar the two are. We use the squared Euclidean distance function to compare normalized molecular spectral signatures, each normalized to have unit peak absorbance. An agglomerative clustering process organizes the pairwise symmetric distance matrix between all pairs of molecules into a cluster hierarchy. Using the hierarchy, one can visualize and identify the molecular spectral features that contribute to the relative detectability of molecules. It is these spectral features that one would want to detect. To locate the distinguishing features, we enumerate the molecules row-by-row in the order of their cluster similarities, imaging a molecule’s spectral signature. The one-dimensional image along the x-axis indicates a bright (yellow) color for high absorbance and dark (blue) for low absorbance (in contrast to a curve plot). Most interesting is the pattern that emerges when one looks at all cluster-enumerated molecular spectral signatures. This phalanx plot shown in Fig. 4 shows that normalized peaks (with some spectral width) at specific wavelengths are shared by many molecules, and these do not contribute significantly to the distinguishability and, thus, relative detectability of the molecule. But it also appears, as shown, that specific clusters share peak distributions that are relatively muted or absent in all the others. These are the distinguishing IR signature features that help detect the molecules in the group. To the degree that the number of molecules in a group is small and the spectral features shared by them are exclusive relative to the others in our All Small Molecule (ASM) database40, they are highly detectable. See88 for more details.

Atmospheric spectra simulator

We simulate model atmospheres to assess the approximate atmospheric abundance of gases needed for detection with simulated James Webb Space Telescope (JWST) observations. We use the computer model “Simulated Exoplanet Atmosphere Spectra” (SEAS) code from88.

We use the molecular mixing ratio profiles and calculate the optical depth of each layer of the atmosphere88,91. We calculate the stellar intensity absorption along each path through the planet atmosphere by A = ni,j σi,j li, where A is absorption, n is number density, σ is the absorption cross-section and l is pathlength. The subscript i denotes each layer that the stellar radiation beam penetrates, and j denotes each molecule. For the height of each atmospheric layer we adopt scale height of the atmosphere. We calculate the transmittance, T, of each beam using the Beer–Lambert Law. Then, we compute the total effective height h of the atmosphere by multiplying the absorption A = 1 − T by the atmosphere's scale height. To connect to observations, we calculate the total attenuated flux as transit depth (Rplanet + h)2/Rstar2 in units of ppm.

We simulate transmission spectra for a hypothetical 1.5 REarth, 5 MEarth super-Earth transiting an M dwarf-star similar to GJ 876. We choose 1.5 REarth as it is consistent with a rocky planet92. We simulate three planetary atmospheres: ones dominated by H2, N2 and CO2. We choose a relatively massive super Earth planet because a more massive planet is more likely to retain an H2-dominated atmosphere than a lower mass planet. We choose to simulate a rocky exoplanet with an H2-dominated atmosphere because such an atmosphere is favorable for detection by transmission spectroscopy than a higher mean molecular weight atmosphere such as one dominated by N2 or CO2. We also simulate an N2 and a CO2-dominated atmosphere. We take the temperature, pressure, and vertical gas abundance profiles from48, where the molecular gas abundances were computed from a photochemical equilibrium model93,94. We assume varying atmospheric abundances of SF6 and NF3, from 1 ppb to 1 ppm.