betaFeedback

WorksPhysics and Astronomy

From Anoxic to Oxic Earth: Detectability of Atmospheric Biosignatures across Stellar Hosts with ELT/ANDES

Abstract

Earth's atmosphere evolved from an anoxic, methane-bearing Archean state through Proterozoic oxygenation to the O2_2-rich modern atmosphere. However, most biosignature detectability studies focus on modern Earth, leaving earlier atmospheric states comparatively unexplored. We investigate how atmospheric evolution and stellar environments influence the detectability of O2_2, CH4_4, H2_2O, and CO2_2, and their potential to distinguish anoxic from oxygenated atmospheres. We construct two complementary photochemical grids: representative Archean, Proterozoic, and modern-Earth scenarios around the Sun, and four geological epochs (3.9, 2.0, 0.8 Ga, and modern Earth) across FGKM hosts. Combining one-dimensional photochemical modeling, high-resolution transmission spectral simulations, and instrument-noise modeling, we assess molecular detectability with ELT/ANDES under optimistic photon-noise-limited conditions. For solar-host Archean and Proterozoic models, CH4_4 requires approximately 33 and 42 transits, respectively, to reach the adopted 5σσ criterion, whereas O2_2 remains below 3σσ within the observing limits. Modern-Earth O2_2 reaches 3σσ after approximately 38 transits. Across the FGKM grid, O2_2 does not reach 5σσ for the 3.9 Ga models. However, For the M3V, K2V, and M8V proxies, modern-Earth O2_2 reaches 5σσ in 9, 15, and 20 transits, respectively, whereas CH4_4 requires 11, 12, and 28 transits at 3.9 Ga and only 2–4 transits across the oxygenated epochs. These results reveal a strong O2_2 detectability contrast around M and K-dwarfs, but not Sun-like hosts in the YJH bands. Robust biosignature interpretation requires complementary molecular constraints, including informative non-detections, within the context of atmospheric evolution and stellar environment.

Introduction

The remote search for life beyond the Solar System is necessarily an exercise in comparative planetology with Earth. Observations of our own planet, from spacecraft flybys to disk-integrated spectra and Earthshine, have demonstrated that atmospheric composition and surface-atmosphere interactions can encode information about habitability and biological activity [67, 71, 38, 59, 66, 65]. However, the modern atmosphere represents only the most recent state in a long evolutionary sequence. Over the past ∼4\sim4 Gyr, Earth evolved from a largely anoxic world through prolonged partial oxygenation to the O2\mathrm{O}_2-rich atmosphere observed today [34, 45, 10, 53]. Evidence for life on Earth extends back to at least ∼3.8\sim3.8 Ga based on isotopic and geological signatures from some of the oldest terrestrial rocks [48, 18], implying that inhabited conditions existed long before modern oxygen levels were established. In the future, temperate rocky exoplanets in habitable zones will be observed at unknown stages of geological and biological evolution, with some of these exoplanets following evolutionary trajectories that only partly resemble Earth’s. Thus, interpreting every potentially habitable planet through a modern-Earth lens could cause us to miss inhabited worlds with little or no oxygen. It could also lead to false positives if abiotic gases are interpreted as signs of life without considering the broader planetary context [47, 69, 77].

The simultaneous presence of an oxidized species such as O2\mathrm{O_2} or O3\mathrm{O_3} and a reduced species such as CH4\mathrm{CH_4} is a long-standing example of atmospheric chemical disequilibrium as a possible biosignature [43, 44, 70, 47, 69]. Its interpretation, however, depends strongly on atmospheric redox state and geochemical boundary conditions [40, 41]. Abiotic processes can generate O2\mathrm{O_2} and O3\mathrm{O_3} under some stellar and atmospheric conditions, CH4\mathrm{CH_4} can have both biological and geological sources, and inhabited planets need not sustain remotely detectable modern-Earth O2\mathrm{O_2} abundances [47, 77]. Individual molecules are therefore insufficient in isolation; O2\mathrm{O_2} and CH4\mathrm{CH_4} are most informative when considered together and with environmental tracers such as CO2\mathrm{CO_2} and H2O\mathrm{H_2O}.

Earth’s own atmospheric history illustrates this problem directly. The Archean atmosphere was predominantly anoxic, with CO2\mathrm{CO_2} and CH4\mathrm{CH_4} being the important greenhouse gases and with organic haze expected under sufficiently methane-rich conditions [78, 1, 10]. Atmospheric O2\mathrm{O_2} began to rise during the Great Oxidation Event near ∼2.4\sim2.4–2.32.3 Ga, but much of the Proterozoic remained weakly oxygenated with atmospheric redox chemistry influenced by CH4\mathrm{CH_4}, volcanic gases, and the sulfur cycle [34, 45, 87, 10]. A later rise in oxygen toward the Neoproterozoic preceded the modern high-O2\mathrm{O_2} atmosphere [45, 53]. These stages span methane-rich anoxic atmospheres, intermediate states containing both reduced and oxygen-bearing species, and the strongly oxidized modern atmosphere. They provide useful evolutionary analogues for inhabited terrestrial planets without implying that all such planets must follow Earth’s specific trajectory.

The observable consequences of these states also depend on the host star. Stellar ultraviolet radiation controls photolysis and therefore the steady-state abundances of species such as O3\mathrm{O_3} and CH4\mathrm{CH_4} [71, 65], while stellar evolution changes the incident radiation field with time [12]. Consequently, similar atmospheric boundary conditions can produce different chemical structures around different stellar hosts. Earth-through-time calculations have explored this dependence from both geological and stellar perspectives. L. Kaltenegger et al. [38] modeled the spectral evolution of Earth across geological epochs, while subsequent studies extended these calculations across FGKM stars [66, 65]. In particular, S. Rugheimer & L. Kaltenegger [64] considered 3.9, 2.0, and 0.8 Ga atmospheres together with modern Earth around an FOV–M8V grid, showing that molecular features depend jointly on geological state and host spectral type. Later high-resolution studies likewise demonstrated that combinations such as O2+CH4\mathrm{O_2+CH_4} and O3+CH4\mathrm{O_3+CH_4} can remain informative before modern oxygen abundances are reached (L. Kaltenegger et al. 2020a,b; R. Payne & L. Kaltenegger 2024).

High-resolution spectroscopy provides a route for testing such atmospheric states observationally. At resolving powers of R∼105R \sim10^{5}, molecular bands are resolved into dense forests of spectral lines whose combined Doppler-shifted signal (at a few km s−1\mathrm{km\,s^{-1}}) can be recovered from quasi-stationary stellar and telluric features through high-resolution cross-correlation spectroscopy (HCCS) [74, 75, 86, 8, 7, 6, 9, 68, 24, 20, 14, 21, 16]. Temperate planets are particularly challenging because their radial velocities change only slowly during transit, increasing the susceptibility of the planetary signal to PCA- or SVD-based detrending. Injection–recovery studies nevertheless show that low-velocity signals can be recovered when sufficient out-of-transit spectra are available and the effect of detrending is propagated to the planetary template [11, 16]. However, stellar line-profile variations, including the Rossiter–McLaughlin effect [46, 61], provide an additional source of contamination because they can also produce phase-dependent residuals that can mimic a slow-moving exoplanetary signal, but whether the most common detrending processes used today for high-resolution spectroscopic analyses can remove this source of contamination still remains poorly understood. Nevertheless, the 39 m Extremely Large Telescope (ELT) and its ArmazoNes high Dispersion Echelle Spectrograph (ANDES) will provide approximately R∼100,000R \sim100{,}000 spectroscopy from optical to near-infrared wavelengths, making temperate rocky planets a major target for high-resolution atmospheric characterization E. Palle et al. [54].

Previous studies illustrate the scale of the observational challenge. I. Snellen et al. [74] obtained a 3.8σ3.8\sigma O2_2 signal after 30 transits for an Earth analogue around an M5V star, while F. Rodler & M. López-Morales [60] estimated approximately 74 transits (∼175\sim175 h) for an M-dwarf system at 10 pc. Using real ground-based spectra, D. B. Serindag & I. A. Snellen [73] estimated about 20–50 transits for an optimal Proxima-Centauri-like case. Molecular detectability can also differ substantially: M. H. Currie et al. [13] found requirements of approximately 34, 33, and 200 transits for CO2_2, CH4_4, and O2_2 detections, respectively, in an Earth-like TRAPPIST-1 e atmosphere. At the population level, K. K. Hardegree-Ullman et al. [29] found that Earth-like O2_2 could be accessible for only up to about 21% of suitable nearby M-dwarf systems over a 50 year survey under their most favorable assumptions. More recently, E. Kurzawa-Ferrandez et al. [42] found H2_2O to be the most accessible species, requiring about 10–19 transits for the TRAPPIST-1 planets and about 30 transits for LHS 1140 b, whereas CO2_2, CH4_4, and O2_2 required approximately 1.5, 3, and 4 times as many transits as H2_2O, respectively. These studies establish the difficulty of high-resolution biosignature searches, but largely consider modern-Earth-like atmospheric compositions.

Here we instead ask how high-resolution molecular detectability changes as an Earth-like atmosphere evolves and as the host star changes. We consider two complementary model grids. The first explores solar-host Archean organic-haze chemistry, Proterozoic O2_2–CH4_4 states, and modern Earth. The second follows four geological epochs across an FGKM host-star grid motivated by S. Rugheimer & L. Kaltenegger [64]. We compute steady-state atmospheric profiles with Photochem N. F. Wogan et al. [85], generate high-resolution transmission spectra with petitRADTRANS P. Mollière et al. [49, 50], and evaluate the detectability of O2_2, CH4_4, H2_2O, and CO2_2 with the ANDES YJH bandpass using the Ratri simulation and cross-correlation framework S. Dash et al. [15, 16]. Our aim is not to assign a unique observable spectrum to a geological age, but to determine how the molecular information available to high-resolution spectroscopy changes with atmospheric

redox state and stellar environment, which combinations of species provide the strongest leverage between broad evolutionary states, and where optimistic photon-noise-limited ANDES predictions indicate the greatest observational potential. § 2 describes the atmospheric and observing framework, § 3 presents the chemical, spectral, and detectability results, § 4 discusses their interpretation and limitations, and § 5 summarizes the implications for future searches for inhabited terrestrial planets that may not resemble modern Earth.

Methods

Atmospheric chemistry models

We use the open-source Photochem package and its EvoAtmosphere solver to simulate the atmospheric chemistry. The code descends from the Atmos/Zahnle photochemical framework and solves coupled gas-phase chemistry, photolysis, transport, condensation, and particle microphysics on a one-dimensional vertical grid [87, 39, 1]. Atmospheric chemistry spans over 100 atmospheric layers in all our models. For each case, the temperature profile and eddy-diffusion profile are prescribed and kept fixed while the chemistry is integrated to steady state. A fully coupled climate-chemistry iteration is beyond the scope of the present study and is deferred to future work.

Archean grid

The Archean models follow the organic-haze (H1) climate families of G. N. Arney et al. [1]. The grid spans four background climate states (A–D) and three CH4/CO2\mathrm{CH}_4/\mathrm{CO}_2 ratios within each state, yielding 12 atmospheres (see Table 1). CO2\mathrm{CO}_2 and CH4\mathrm{CH}_4 are imposed through fixed lower-boundary partial pressures, O2\mathrm{O}_2 is fixed at 10−810^{-8} bar, and N2\mathrm{N}_2 fills the remaining dry atmosphere. We use the chemical network zahnle_HNOCS.yaml, which includes hydrocarbon chemistry and organic-haze precursors, for our study. Following Supplementary Table 2 of G. N. Arney et al. [1], lower boundary conditions combine surface deposition with volcanic outgassing, where both apply. H2\mathrm{H}_2, SO2\mathrm{SO}_2, and H2S\mathrm{H}_2\mathrm{S} are supplied at volcanic fluxes of 3.5×1093.5 \times10^9, 101010^{10}, and 3.5×1083.5 \times10^8 molecules cm−2s−1\mathrm{cm}^{-2}\mathrm{s}^{-1} together with deposition velocities of 2.4×10−42.4 \times10^{-4}, 1.01.0, and 0.0210.021 cm s−1\mathrm{cm}\ \mathrm{s}^{-1}, respectively. CO is removed by deposition alone at 1.2×10−41.2 \times10^{-4} cm s−1\mathrm{cm}\ \mathrm{s}^{-1}. Additional deposition velocities are imposed for O, H, OH, HO2\mathrm{HO}_2, HCO, CH3\mathrm{CH}_3, HNO, HSO, H2SO4\mathrm{H}_2\mathrm{SO}_4, H2O2\mathrm{H}_2\mathrm{O}_2, H2CO\mathrm{H}_2\mathrm{CO}, HNO3\mathrm{HNO}_3, NO, NO2\mathrm{NO}_2, O3\mathrm{O}_3 and the condensed particles S8\mathrm{S}_8, H2SO4\mathrm{H}_2\mathrm{SO}_4, and HCaer1 (for further details, see Supplementary tbl:S2 from G. N. Arney et al. [1]). Water vapor is maintained with an ocean reservoir and is allowed to condense onto H2Oaer\mathrm{H}_2\mathrm{O}_{\mathrm{aer}} at a relative humidity of 0.8. Case-to-case changes are restricted to the fixed surface pressures of N2\mathrm{N}_2, CO2\mathrm{CO}_2, and CH4\mathrm{CH}_4 (with O2\mathrm{O}_2 held at 10−810^{-8} bar in all Archean H1 runs). The initial pressure–temperature profiles are generated with AdiabatClimate (N. F. Wogan et al. [85]), using the haze-bearing surface temperatures reported by G. N. Arney et al. [1], an isothermal stratosphere at 180 K, and Kzz=106K_{zz} = 10^6 cm2s−1\mathrm{cm}^2\mathrm{s}^{-1}.

CasePsurfP_{\mathrm{surf}} (bar)pCO2p\mathrm{CO}_2 (bar)CH4_4/CO2_2pCH4p\mathrm{CH}_4 (bar)TsurfT_{\mathrm{surf}} (K)
A1-H11.00.01000.150.00150284
A2-H11.00.01000.190.00190263
A3-H11.00.01000.260.00260263
B1-H11.00.01800.100.00180299
B2-H11.00.01800.170.00306274
B3-H11.00.01800.210.00378274
C1-H10.50.01000.150.00150282
C2-H10.50.01000.180.00180282
C3-H10.50.01000.210.00210262
D1-H10.50.00360.200.00072273
D2-H10.50.00360.280.001008251
D3-H10.50.00360.310.001116251

Table 1. Table 1. Archean H1 photochemical grid following G. N. Arney et al. [1]. All cases use a constant pO2=10−8p\mathrm{O}_2 = 10^{-8} bar.

Casesf(O2)f(\mathrm{O}_2)f(CH4)f(\mathrm{CH}_4)Notes
A1-V210−1010^{-10}10−610^{-6}very anoxic, low CH4\mathrm{CH}_4
A2-V210−1010^{-10}10−510^{-5}
A3-V210−1010^{-10}10−410^{-4}
A4-V210−1010^{-10}10−310^{-3}very anoxic, CH4\mathrm{CH}_4-rich
B1-V210−710^{-7}10−610^{-6}weakly oxygenated, low CH4\mathrm{CH}_4
B2-V210−710^{-7}10−510^{-5}
B3-V210−710^{-7}10−410^{-4}
B4-V210−710^{-7}10−310^{-3}weakly oxygenated, CH4\mathrm{CH}_4-rich
C1-V210−310^{-3}10−610^{-6}oxic, low CH4\mathrm{CH}_4
C2-V210−310^{-3}10−510^{-5}
C3-V210−310^{-3}10−410^{-4}
C4-V210−310^{-3}10−310^{-3}oxic, CH4\mathrm{CH}_4-rich

Table 2. Proterozoic V2 photochemical grid from K. Zahnle et al. [87]. The background atmosphere has pCO2=0.01p\mathrm{CO}_2 = 0.01 bar and Tsurf=284T_{\mathrm{surf}} = 284 K.

The incident spectrum is a 2.7 Ga solar spectrum based on the wavelength-dependent solar evolution prescription of M. W. Claire et al. [12]. This age follows the late-Archean framework of G. N. Arney et al. [1], who adopted a 2.7 Ga Sun to match the age of the geochemical CO2\mathrm{CO}_2 constraints used to define their atmospheric scenarios (S. G. Driese et al. [19]). It should therefore be regarded as representative of a late-Archean atmosphere. Because solar evolution is wavelength dependent, an earlier Archean Sun would differ most strongly at short UV wavelengths relevant to photolysis and haze-producing chemistry; exploring this temporal dependence across the full Archean is beyond the scope of the present grid. We use a surface albedo of 0.32, a solar zenith angle of 60∘60^\circ, and diffusion-limited hydrogen escape. The resulting grid is intended to follow the transition from comparatively haze-poor to strongly methane-driven hydrocarbon chemistry while retaining the background climate states explored by G. N. Arney et al. [1].

Proterozoic grid

The Proterozoic models are constructed around the two principal redox controls explored by K. Zahnle et al. [87]: surface O2\mathrm{O}_2 and CH4\mathrm{CH}_4. We use a 1 bar N2\mathrm{N}_2 background with pCO2=0.01p\mathrm{CO}_2 = 0.01 bar and the “modern high” (V2) volcanic sulfur fluxes. Surface O2\mathrm{O}_2 mixing ratios are 10−1010^{-10}, 10−710^{-7}, and 10−310^{-3}, while CH4\mathrm{CH}_4 mixing ratios are 10−610^{-6}, 10−510^{-5}, 10−410^{-4}, and 10−310^{-3}. The resulting 12 models are listed in Table 2. Organic haze is disabled so that changes in the gas-phase oxidation state can be isolated.

The thermal structure of the atmosphere follows the representative setup of K. Zahnle et al. [87], with Tsurf=284T_{\mathrm{surf}} = 284 K, a tropopause near 11 km, and an isothermal 186 K stratosphere. The eddy diffusion coefficient is 10510^5 cm2s−1\mathrm{cm}^2\mathrm{s}^{-1} in the troposphere and decreases with pressure in the stratosphere following the Zahnle prescription. We use a solar zenith angle of 50∘50^\circ, a surface albedo of 0.32, and relative humidity of 0.8. The stellar spectrum represents a 2.3 Ga Sun with the ultraviolet enhancement adopted by K. Zahnle et al. [87]: a factor of four shortwards of 175 nm and two from 175–400 nm relative to the modern spectrum, together with the solar-age scaling of M. W. Claire et al. [12] and I. Ribas et al. [57].

Modern Earth benchmark

The modern benchmark uses the zahnle_earth.yaml network and the present-day solar spectrum distributed with Photochem (the comparison of the solar spectrum across three ages are shown in Figure 1). The lower boundary fixes pN2=0.78p_{\mathrm{N}_2}=0.78 bar, pO2=0.21p_{\mathrm{O}_2}=0.21 bar, pCO2=3.6×10−4p_{\mathrm{CO}_2}=3.6 \times10^{-4} bar, and trace abundances of H2\mathrm{H}_2, CO, and N2O\mathrm{N}_2\mathrm{O}, while CH4\mathrm{CH}_4 is supplied at 9×10109 \times10^{10} molecules cm−2^{-2} s−1^{-1}. SO2\mathrm{SO}_2 and H2S\mathrm{H}_2\mathrm{S} are included at fluxes of 10910^9 and 10810^8 molecules cm−2^{-2} s−1^{-1}, respectively. Rainout and species-dependent dry deposition are included for soluble and reactive species. We keep a solar zenith angle of 60∘60^\circ, a surface albedo of 0.25, a near-surface Kzz=105K_{zz}=10^5 cm2^2 s−1^{-1}, and water condensation at a relative humidity of 0.5. The converged surface temperature is approximately 286.5 K.

Figure 1. Solar ultraviolet spectra implemented for the Archean, Proterozoic, and modern solar-host calculations. The curves follow the separate stellar prescriptions used for the three model families described in § 2.1.1 and § 2.1.2.

FGKM Earth-through-time grid

To examine the effect of the stellar environment, we construct a second atmospheric grid following the design of S. Rugheimer & L. Kaltenegger [64]. Four epochs are considered across stellar hosts: 3.9 Ga, 2.0 Ga, 0.8 Ga, and modern Earth. The host grid contains F0V, F7V, G8V, K2V, K7V, M1V, M3V, and M8V spectral types, together with GJ 581, GJ 832, and GJ 1214, yielding 44 atmosphere-host combinations. The proxy star does not necessarily belong to the same nominal spectral subtype for all the host star cases, because suitable UV spectra are not available for every spectral type in the MUSCLES databases. Where an exact match is unavailable, we use the closest suitable stellar proxy with available UV coverage.

At a given epoch, each host delivers the same bolometric irradiation to the model atmosphere, as in S. Rugheimer & L. Kaltenegger [64]. The top-of-atmosphere flux is

F⋆=1370fL W m−2,(1)F_\star= 1370 f_L\ \mathrm{W\,m^{-2}}, \tag*{(1)}

where fL=0.746f_L=0.746, 0.852, 0.935, and 1.0 for 3.9 Ga, 2.0 Ga, 0.8 Ga, and modern Earth, respectively (M. W. Claire et al. [12]). The wavelength dependence of the ultraviolet radiation is taken from MUSCLES, Mega-MUSCLES, or MUSCLES Extension proxy spectra (K. France et al. [22]; D. J. Wilson et al. [82]; P. R. Behr et al. [4]; D. J. Wilson et al. [83]). The proxy spectra are renormalized to the appropriate epoch-integrated flux before being supplied to Photochem. The final host-star UV spectra are shown in Figure 2.

Figure 2. Ultraviolet spectral energy distributions for the multi-star grid. Each stellar proxy is renormalized to the epoch-dependent bolometric irradiation used by the photochemical model. The wavelength-dependent UV shape therefore varies with host star and directly controls photolysis.

Surface CO2_2 and O2_2 are fixed to the epoch-dependent values of S. Rugheimer & L. Kaltenegger [64], while the CH4_4 and N2_2O lower-boundary abundances follow their host-dependent values. Surface temperatures TsurfT_{\mathrm{surf}} for the F0V-M8V grid points are taken from the top panel of Figure 2 of S. Rugheimer & L. Kaltenegger [64]. For the additional MUSCLES hosts GJ 581, GJ 832, and GJ 1214, we assign TsurfT_{\mathrm{surf}} at each epoch by linear interpolation in effective temperature between the M3V host (Teff=3200T_{\mathrm{eff}}=3200 K) and M1V host (Teff=3700T_{\mathrm{eff}}=3700 K) anchors,

Tsurf(Teff)=TsurfM3V+Teff−3200 K500 K(TsurfM1V−TsurfM3V),(2)T_{\mathrm{surf}}(T_{\mathrm{eff}})=T_{\mathrm{surf}}^{\mathrm{M3V}}+\frac{T_{\mathrm{eff}}-3200\ \mathrm{K}}{500\ \mathrm{K}}\left(T_{\mathrm{surf}}^{\mathrm{M1V}}-T_{\mathrm{surf}}^{\mathrm{M3V}}\right), \tag*{(2)}

and round the result to the nearest kelvin. The host effective temperatures used for this interpolation are Teff=3498T_{\mathrm{eff}}=3498 K (GJ 581), 3620 K (GJ 832), and 3250 K (GJ 1214). The models use a 1 bar dry atmosphere, Earth mass and radius, a surface albedo of 0.32, a solar zenith angle of 60∘60^\circ, and Kzz=105K_{zz}=10^5 cm2^2 s−1^{-1}. The 3.9 Ga cases use the H-N-O-C-S network, whereas the oxygenated epochs use the modern-Earth network. We do not impose organic haze in this grid.

High-resolution transmission spectra

For each converged atmospheric column, we compute line-by-line radiative transfer and obtain clear-atmosphere transmission spectra with petitRADTRANS. The pressure, temperature, mean molecular weight, and volume mixing-ratio profiles from Photochem are mapped onto the radiative-transfer grid. All planets are assigned Rp=1R⊕R_p=1R_\oplus and Mp=1M⊕M_p=1M_\oplus, with a reference radius specified at 0.01 bar. Spectra are calculated from 0.4–1.8 μm at an effective resolving power of 10510^5.

The opacity budget includes CO2_2, CH4_4, CO, HCN, C2_2H6_6, H2_2O, O2_2, and O3_3 where present; the multi-star grid also includes N2_2O. In all simulations we include Rayleigh scattering by N2_2, O2_2, H2_2O, CO2_2, and H2_2 whenever those species are present in the Photochem abundance profiles. N2_2 is the dominant background gas in these Earth-like atmospheres, so N2_2 Rayleigh sets most of the continuum baseline; the remaining Rayleigh partners are retained because O2_2, H2_2O, and CO2_2 can contribute non-negligibly when their abundances rise (oxygenated eras; moist or CO2_2-rich cases), while H2_2 Rayleigh is included for consistency in H2_2-bearing Archean mixtures. The adopted line data are based on HITEMP and HITRAN databases: H2_2O, CO2_2, and CO from HITEMP (L. S. Rothman et al. [62]); CH4_4 from R. J. Hargreaves et al. [30]; HCN from

Note—a^{a}The exposure-time floor places these nearby systems near the soft-saturation regime.

Note—b^{b}The MUSCLES radius is unavailable; we took R⋆=0.71R⊙R_{\star} = 0.71R_{\odot} from K. G. Stassun et al. [76].

Note—c^{c}The solar-host atmospheric calculations use the epoch-dependent solar spectra described in § 2.1.1–§ 2.1.3.

G. J. Harris et al. [33]; C2_2H6_6 uses HITRAN2024 I. E. Gordon et al. [26], and N2_2 from the WC-CRMT database C. M. Western et al. [80]. The high-resolution O2_2 I. E. Gordon et al. [25], O3_3 L. S. Rothman et al. [63], and N2_2O R. J. Hargreaves et al. [31] tables were obtained from DACE opacity database S. L. Grimm et al. [28] and converted into the petitRADTRANS format. The O2_2 opacity includes the 0.76 μ\mum A-band, whereas our O3_3 table begins at 1.4 μ\mum. Organic haze and clouds are intentionally omitted from the radiative transfer. We also neglect atmospheric

refraction for our temperate Earth-twin transit geometry; under these conditions gas and Rayleigh opacity dominate the slant optical depth, so refraction remains as a minor contributor to the continuum [58, 5]. The resulting spectra should therefore be interpreted as gas-only limiting cases for HRCSS. Haze is retained in the Archean photochemistry because it affects the chemical state of the atmosphere, but its continuum extinction is not propagated into the transmission spectrum.

ELT/ANDES simulations with Ratri

The generated high-resolution transmission spectra are used as inputs to the open-source Ratri1 high-resolution observational simulation framework to estimate their detectabilities using ELT/ANDES. Ratri itself has already been validated against the current version of the ANDES ETC simulator across all bands in S. Dash et al. [16] for its current configuration, but in this study we retain only the YJH band with CARMENES CCD configuration used in S. Dash et al. [16] and generate a noisy exposure (with a corresponding signal-to-noise per pixel or S/Nnspectra=1S/N_{n_{\mathrm{spectra}}=1}; see Figure 20) incorporating wavelength-dependent stellar photon flux, telluric absorption through the Earth’s atmosphere, instrumental response tailored to CARMENES CCDs from S. Gandhi et al. [23], and a Poissonian noise treatment for calculating the signal-to-noise (S/NS/N) per pixel. The errors at all pixels are assumed to be uncorrelated, and thus the actual noise at each pixel is randomly drawn from a normal distribution with variance equivalent to the photon flux error value calculated at that pixel assuming Poisson statistics. Since we anticipate the difficulty of detecting our generated spectra for almost all stellar hosts within a few transits, we use the time-invariant form of Ratri using the ratri_1D.txt script. To simulate telluric absorption, we use an airmass value of 1.2 and a precipitable water vapor (PWV) value between 2.5–3.5 mm as inputs for the ESO SkyCalc module (S. Noll et al. [52]; A. Jones et al. [35]) integrated within Ratri.

Figure 20. Continuum signal-to-noise ratio per pixel as a function of wavelength for the ANDES YJH mode at each host’s recommended exposure time (texpt_{\mathrm{exp}}). Exposures are auto-scaled so the stellar continuum reaches an assumed unsaturated level of ∼7×104 e−\sim7 \times10^{4}\,\mathrm{e}^{-}. Most hosts sit in that regime; the K2V proxy (ε\varepsilon Eri) remains soft-saturated even at the 0.5 s floor, and GJ 832 at the same floor lies near soft saturation. Toward cooler hosts, the relative S/NS/N shifts toward longer wavelengths within the YJH range, with the H band becoming increasingly favorable relative to the Y and J bands.

The time-invariant script injects the exoplanetary signal at one particular value of exoplanetary radial velocity (RV) for an exposure (away from 0 km s−1^{-1} where any residual telluric contamination can artificially enhance the signal due to artifacting), and then stacks nspectran_{\mathrm{spectra}} exposures together to enhance the total S/NS/N ratio per pixel through the relation:

(SN)nspectra=nspectra(SN)nspectra=1=Snspectra=1Nnspectra=1/nspectra.(3)\left(\frac{S}{N}\right)_{n_{\mathrm{spectra}}} = \sqrt{n_{\mathrm{spectra}}}\left(\frac{S}{N}\right)_{n_{\mathrm{spectra}}=1} = \frac{S_{n_{\mathrm{spectra}}=1}}{N_{n_{\mathrm{spectra}}=1}/\sqrt{n_{\mathrm{spectra}}}.} \tag*{(3)}

The above relation follows from the fact that the total signal ∝nspectra\propto n_{\mathrm{spectra}}, with the total noise ∝nspectra\propto\sqrt{n_{\mathrm{spectra}}} assuming Poisson statistics. The factor of nspectra\sqrt{n_{\mathrm{spectra}}} used to damp down the noise for a single exposure in the last term of the equation is referred to as the damping factor, which carries information about the total number of exposures stacked to get the corresponding S/NS/N.

After damping down the noise, we perform an idealized detrending step. The noisy exposure is divided by a simulation of the same exposure without noise or an injected planetary signal. This leaves a residual array containing only the noise and the injected planetary signal at a given RV. Here, we also mask the pixels S/N≤50S/N \le50 as a low-flux correction step to remove the contribution of pixels that would realistically have the most telluric contamination and would also be masked out in real observations as well. Subsequently, a search for the injected signal is carried out by Doppler-shifting a model spectral template across a grid of RVs (on to the same pixel array as the residual dataset) and then cross-correlating with the residual array, with the same pixels also masked for the Doppler-shifted model array as in the residual array calculated before. If the peak of the cross-correlation function (CCF) stands out appreciably at some value of RV in comparison to its neighbouring values, we count it as a detection. To quantify the significance of detection of the peak of CCF compared to the rest of the RV-grid after cross-correlation, we use the CCF-to-log LL (log-Likelihood) equation of M. Brogi & M. R. Line [9]:

log⁡(L)=−N2log⁡[sf2−2R(s)+sg2],(4)\log(L) = -\frac{N}{2}\log\left[s_f^2 - 2R(s) + s_g^2\right], \tag*{(4)}

where sf2s_f^2 (data variance), sg2s_g^2 (model variance) and R(s)R(s) (cross-covariance) are calculated as

sf2=1N∑nf2(n),s_f^2 = \frac{1}{N}\sum_n f^2(n),
sg2=1N∑ng2(n),s_g^2 = \frac{1}{N}\sum_n g^2(n),
R(s)=1N∑nf(n)g(n−s).(5)R(s) = \frac{1}{N}\sum_n f(n)g(n-s). \tag*{(5)}

f(n)f(n) and g(n)g(n) are found by subtracting the mean from all unmasked values in each row of each spectroscopic order in the detrended residual array, and the corresponding row from the Doppler-shifted model array respectively. From here, we use a 1D version (with RV as the sole free-parameter) of the σ\sigma-calculation criterion based on Wilks’ theorem (S. S. Wilks [81]) used in S. Dash et al. [14]. We note that this σ\sigma calculation is based on likelihoods and is different from the CCF-based signal-to-noise ratio (SNR) metric used in many other literature. For our metric, a signal confidently detected will have a value of 0–σ\sigma and the rest of the grid will have values >4−5σ> 4 - 5\sigma, thus meaning that our analysis is able to confidently reject most of the tested RV-grid at >4−5σ> 4 - 5\sigma compared to the injected RV solution.

Our detection metric differs from the evidence-based approach of E. Kurzawa-Fernandez et al. (2026). They compute Bayesian evidences, log⁡(Z)\log(Z), using Nested Sampling and compare a signal model, in which the data are cross-correlated with Doppler-shifted planetary templates, against a null model (no injected planetary signal) through a Bayes factor test. Our approach thus tests how significant the signal at the RV value corresponding to the CCF peak is compared with the CCF values across the rest of the RV grid. Rather than testing whether the signal itself can be distinguished from the cross-correlation obtained from a dataset without any injected signal, it evaluates the significance of the recovered peak relative to the surrounding RV grid. However, we note that spurious residuals mimicking a signal rarely go above a 3σ3\sigma level with our approach and generally lie away from the injected RV values (see the bottom panel in Figure 11 for the 3.9 Ga case). Hence, we prefer to use a much tighter 5σ5\sigma threshold to count true detections to limit the possibility of a spurious signal leading to false detections of molecules for our grid of models.

Figure 11. Stacked CH4_4 CCFs for the M8V (TRAPPIST-1) proxy at the four geological epochs using 225 frames (texp ≈\approx 90.5 s, airmass = 1.2). (Top): CCF versus RV shift; the dashed vertical line marks the time-invariant injected systemic radial velocity (−20-20 km s−1^{-1}). (Bottom): significance with which each trial RV can be statistically excluded as the planet’s true velocity; the curve dips sharply to its minimum at the injected RV with 3σ3\sigma and 5σ5\sigma thresholds marked for reference (dashed horizontal lines). The depth of this dip relative to the surrounding curve gives the detection significance for each epoch: 3.9 Ga (2.83σ\sigma - a non-detection), 2.0 Ga (8.22σ\sigma), 0.8 Ga (9.79σ\sigma), modern (7.62σ\sigma).

For the solar-host detectability calculations, we use a Sun-like star with Teff=5772 KT_{\mathrm{eff}} = 5772\ \mathrm{K}, R⋆=1R⊙R_{\star} = 1R_{\odot}, and M⋆=1M⊙M_{\star} = 1M_{\odot}. The system is placed at a fiducial distance of 150 pc, for which an exposure time of 30 s keeps the continuum count level close to 7×1047 \times10^{4} electrons per resolution element while avoiding saturation of the CCD pixels. This operating point remains below the adopted soft-saturation limit and is conservative relative to the >70000 e−>70000\ e^{-} per-pixel well capacity reported for the Teledyne HAWAII-4RG-15 (2.5 μ\mum cutoff) detectors baselined for ANDES YJH ([88]). To estimate the number of spectra that can be obtained within a given observing interval, and hence the value of nspectran_{\mathrm{spectra}} used in Equation (3), we assume a 34 s readout time between successive exposures, following the CARMENES timing currently implemented in Ratri ([56, 16]). This is slightly more conservative than the 30 s overhead stated by E. Palle et al. [54]. Using the same Earth-twin orbital prescription and impact parameter of b=0.3b = 0.3 described in § 2.3.1, Equation (11) gives a transit duration of T14≃12.5 hT_{14} \simeq12.5\ \mathrm{h}. This duration is used to convert the required number of simulated frames into transits and equivalent observing nights.

For each stellar proxy in the FGKM grid, the exposure time is similarly adjusted to keep the continuum count level close to 7×1047 \times10^{4} electrons per resolution element while avoiding saturation of the CCD pixels, subject to a minimum exposure time of 0.5 s. For the nearest K- and M-dwarf proxies, the 0.5 s exposure-time floor places some simulations close to the soft-saturation regime, and the corresponding detectability estimates should therefore be interpreted with caution.

To enable comparisons of detectability, we standardize the calculated sigma to that obtained from a stacking of Nref=1600N_{\mathrm{ref}} = 1600 simulated frames. For each atmosphere–host-star combination, we measure the corresponding detection significance, σref\sigma_{\mathrm{ref}} at Nref=1600N_{\mathrm{ref}} = 1600. In the photon-noise-dominated regime, this significance is expected to scale approximately as N\sqrt{N} with the number of combined exposures ([74, 6]). We therefore estimate the number of frames (NframesN_{\mathrm{frames}}) required to reach a target significance (σt\sigma_t) as

Nframes=Nref(σtσref)2.(6)N_{\mathrm{frames}} = N_{\mathrm{ref}}\left(\frac{\sigma_t}{\sigma_{\mathrm{ref}}}\right)^2. \tag*{(6)}

Unless stated otherwise, we adopt σt=5\sigma_t = 5 to correspond to a strong detection. We do not impose a minimum significance threshold on the 1600-frame reference calculation. For every finite and positive value of σref\sigma_{\mathrm{ref}}, the number of frames required to reach the target significance is estimated using the N\sqrt{N} scaling described above and converted into the corresponding number of transits and equivalent observing nights. Configurations requiring more than 500 transits or 200 equivalent observing nights are considered not practically detectable within the observing limits and are denoted as n.d. (non-detection). Cases with missing, non-finite, or non-positive values of σref\sigma_{\mathrm{ref}} are treated similarly.

Orbital and observing assumptions

For the multi-star grid, each atmospheric model is assigned to an Earth twin with Rp=1R⊕R_p = 1R_{\oplus} and Mp=1M⊕M_p = 1M_{\oplus} on a circular orbit. The planet is placed at an orbital distance where the total incident bolometric flux equals that received by present-day Earth. This choice provides a common irradiation reference across the stellar grid and is not intended to define the planet as lying within the habitable zone, although the resulting orbit may overlap with conventional habitable-zone limits for some stellar types. The stellar luminosity (L⋆L_{\star}) is calculated from the proxy radius (R⋆R_{\star}) and effective temperature (TeffT_{\mathrm{eff}}),

L⋆L⊙=(R⋆R⊙)2(TeffTeff,⊙)4,(7)\frac{L_{\star}}{L_{\odot}} = \left(\frac{R_{\star}}{R_{\odot}}\right)^2 \left(\frac{T_{\mathrm{eff}}}{T_{\mathrm{eff},\odot}}\right)^4, \tag*{(7)}

where Teff,⊙=5772 KT_{\mathrm{eff},\odot} = 5772\ \mathrm{K}. The stellar mass (M⋆M_{\star}) is estimated from the radius and surface gravity (log⁡g\log g) used in the simulation,

M⋆M⊙=(R⋆R⊙)210log⁡g−log⁡g⊙,(8)\frac{M_{\star}}{M_{\odot}} = \left(\frac{R_{\star}}{R_{\odot}}\right)^2 10^{\log g-\log g_{\odot}}, \tag*{(8)}

with log⁡g⊙=4.437\log g_{\odot} = 4.437. The same stellar radius is used when calculating the transit depth through (Rp/R⋆)2(R_p/R_{\star})^2. The Earth-equivalent semi-major axis is then

aAU=L⋆L⊙(9)\frac{a}{\mathrm{AU}} = \sqrt{\frac{L_{\star}}{L_{\odot}}} \tag*{(9)}

and the orbital period follows from Kepler’s third law,

(P1 yr)2=(a/AU)3M⋆/M⊙.(10)\left(\frac{P}{1\ \mathrm{yr}}\right)^2 = \frac{(a/\mathrm{AU})^3}{M_{\star}/M_{\odot}}. \tag*{(10)}

This recipe provides a common Earth-twin geometry across the stellar grid rather than reproducing the orbital properties of any known planet, and is consistent in spirit with placing the planet at Earth-like irradiation within the classical habitable-zone framework (R. K. Kopparapu et al. [39]). The orbital configuration and the epoch-dependent irradiation used in the atmospheric calculations are treated separately. For each stellar proxy, the semi-major axis is fixed using its present-day luminosity and is kept unchanged across all four geological epochs. The luminosity scaling factors fLf_L obtained

from the Earth-through-time framework (M. W. Claire et al. [12]; S. Rugheimer & L. Kaltenegger [64]) are instead applied to the stellar spectrum supplied to the photochemical calculation. Thus, the epoch-dependent scaling modifies the irradiation driving the atmospheric chemistry and thermal structure, while the orbital distance, period, and transit geometry remain fixed for a given host star.

We assume a fixed impact parameter of b=0.3b = 0.3 for all systems. For a circular orbit, the transit duration of a planet is calculated following J. N. Winn [84] as

T14=Pπarcsin⁡[R⋆a(1+Rp/R⋆)2−b21−(bR⋆/a)2].(11)T_{14} = \frac{P}{\pi}\arcsin\left[\frac{R_{\star}}{a}\sqrt{\frac{(1+R_{\mathrm{p}}/R_{\star})^2-b^2}{1-(bR_{\star}/a)^2}}\right]. \tag*{(11)}

Each simulated transit consists of an in-transit interval of duration T14T_{14} together with an out-of-transit baseline of equal duration. In theory, this is a conservative estimate, since partial-transit observations with sufficient out-of-transit baseline may also yield a detection. The effective cadence is

tcad=texp+tduty,(12)t_{\mathrm{cad}} = t_{\mathrm{exp}} + t_{\mathrm{duty}}, \tag*{(12)}

where tduty=34t_{\mathrm{duty}} = 34 s. The number of spectra obtained per transit event is therefore

Nper transit=2T14tcad,(13)N_{\mathrm{per\ transit}} = \frac{2T_{14}}{t_{\mathrm{cad}}}, \tag*{(13)}

with T14T_{14} and tcadt_{\mathrm{cad}} expressed in same units. The required number of transit observations is then

Ntransits=NframesNper transit(14)N_{\mathrm{transits}} = \frac{N_{\mathrm{frames}}}{N_{\mathrm{per\ transit}}} \tag*{(14)}

To facilitate comparison between systems with substantially different transit durations, we additionally express the observing requirement in equivalent 7 h units of on-sky telescope time following E. Kurzawa-Ferrandez et al. [42],

Nnights=[2NtransitsT147 h](15)N_{\mathrm{nights}} = \left[\frac{2N_{\mathrm{transits}}T_{14}}{7\ \mathrm{h}}\right] \tag*{(15)}

Here, NnightsN_{\mathrm{nights}} represents the total on-sky observing time expressed in 7 h units and should not be interpreted as the number of calendar nights required to complete the observations. Actual scheduling feasibility will additionally depend on transit ephemerides, seasonal visibility, weather losses, and other operational constraints. Table 3 summarizes the stellar and orbital properties used for the detectability calculations. The distance to each stellar proxy determines the photon budget, while the adopted stellar radius sets the transmission contrast.

HostProxyR⋆R_{\star} (R⊙R_{\odot})dd (pc)PP (days)T14T_{14} (hours)texpt_{\mathrm{exp}} (s)SED source
F0VWASP-171.57403898.223.8169.0P. R. Behr et al. [4]
F7VWASP-171.57403898.223.8169.0P. R. Behr et al. [4]
Sun–1.00150365.2512.5030.0Solarc^{c}
G8VTOI-1930.9580.44293.311.1410.5P. R. Behr et al. [4]
K2Vε\varepsilon Eri0.763.212175.77.970.5a^{a}K. France et al. [22]
K7VHD 855120.71b^{b}11.28124.27.170.5K. France et al. [22]
M1VGJ 1760.489.48545.03.791.4K. France et al. [22]
M3VGJ 6990.181.82715.41.800.55D. J. Wilson et al. [83]
M8VTRAPPIST-10.1212.474.410.8290.5D. J. Wilson et al. [82]
GJ 581GJ 5810.326.30025.62.451.6K. France et al. [22]
GJ 832GJ 8320.484.96747.54.100.5a^{a}K. France et al. [22]
GJ 1214GJ 12140.2214.6413.41.6525.5K. France et al. [22]

Table 3. Stellar hosts used for the detectability calculations. Period (PP) and transit duration (T14T_{14}) correspond to an Earth twin on a circular Earth-insolation orbit with impact parameter b=0.3b = 0.3. Exposure times are those used in the Ratri simulations for each stellar host.

Results

Photochemical evolution from Archean to modern conditions

The Archean calculations show the expected transition from weakly reducing to methane-rich, haze-forming chemistry (see Figure 3). N2\mathrm{N}_2 is treated as the dominant background atmospheric constituent and is therefore not shown in the figure, while the imposed CH4/CO2\mathrm{CH}_4/\mathrm{CO}_2 ratio controls the reduced-carbon inventory. Increasing CH4/CO2\mathrm{CH}_4/\mathrm{CO}_2 raises the CH4\mathrm{CH}_4 abundance throughout the column and enhances secondary hydrocarbons, including C2H6\mathrm{C}_2\mathrm{H}_6 and haze precursors. The strongest haze production occurs in the high-CH4/CO2\mathrm{CH}_4/\mathrm{CO}_2 members of each climate family, broadly consistent with the behavior identified by G. N. Arney et al. [1] and laboratory haze experiments [78]. O2\mathrm{O}_2 remains fixed near the anoxic lower-boundary value and O3\mathrm{O}_3 is negligible. Thus, within this grid, methane and its photochemical products provide the dominant spectroscopic distinction among the Archean models.

Figure 3. Steady-state mixing-ratio profiles for the Archean H1 grid. The four panels correspond to the A–D climate families from Table 1; within each panel, increasing CH4/CO2\mathrm{CH_4}/\mathrm{CO_2} drives progressively stronger hydrocarbon chemistry.

The Proterozoic grid separates the effects of atmospheric oxidation and methane abundance (see Figure 4). At pO2=10−10p\mathrm{O}_2 = 10^{-10} bar, the column remains strongly anoxic and the CH4\mathrm{CH}_4 profiles largely track the imposed lower boundary. Raising O2\mathrm{O}_2 to 10−710^{-7} bar produces an intermediate redox state in which substantial reduced carbon can coexist with a larger oxidizing reservoir. At 10−310^{-3} bar, O2\mathrm{O}_2 becomes a major constituent and an O3\mathrm{O}_3 layer develops, particularly in the methane-poor models.

Figure 4. Steady-state mixing-ratio profiles for the Proterozoic grid at surface O2_2 mixing ratios of 10−1010^{-10}, 10−710^{-7}, and 10−310^{-3}. Within each panel, CH4_4 increases from 10−610^{-6} to 10−310^{-3} at the lower boundary.

The modern Earth benchmark (Figure 5) is qualitatively distinct from both earlier grids: O2\mathrm{O}_2 dominates the oxidizing inventory, the stratospheric O3\mathrm{O}_3 layer is well developed, and CH4\mathrm{CH}_4 is strongly suppressed relative to the high-methane Archean and Proterozoic cases. This atmospheric state provides the conventional reference for comparison with the earlier Archean- and Proterozoic-like stages of Earth’s atmospheric evolution considered here.

Figure 5. Steady-state mixing-ratio structure of the modern Earth benchmark. The atmosphere is O2_2 rich, contains a stratospheric O3_3 layer, and retains only trace CH4_4 relative to the Archean and Proterozoic methane-rich cases.

The multi-star grid retains the epoch ordering imposed by the lower-boundary conditions while introducing a strong host-star dependence (the mixing ratio profiles for all species across stellar hosts are shown in Figure 6). The 3.9 Ga atmospheres are effectively anoxic for all hosts. Oxygen increases at 2.0 Ga and 0.8 Ga and reaches the modern value in the final epoch. At fixed oxygenated epoch, CH4\mathrm{CH}_4 is generally enhanced around cooler hosts, consistent with previous calculations showing that the UV environments of K and M dwarfs can lengthen the atmospheric lifetime of reduced species [71, 65, 64]. The host spectrum therefore modulates the detailed chemical state without erasing the broad evolution from anoxic to oxic conditions.

Figure 6. Representative mixing-ratio profiles for the FGKM host × geological-epoch grid following S. Rugheimer & L. Kaltenegger [64]. The dominant temporal trend is the rise of O2_2 and O3_3 with geological age, while host-star UV irradiation alters the abundance and vertical distribution of photochemically sensitive species such as CH4_4.

High-resolution spectral evolution

Figure 7 summarizes the principal high-resolution spectral changes for the Sun-Earth system over geological timescale. The Archean spectrum is methane dominated, with small additional features from H2O\mathrm{H}_2\mathrm{O} and CO2\mathrm{CO}_2. The selected Proterozoic model develops strong O2\mathrm{O}_2 absorption while retaining a chemically distinct reduced component. Modern Earth produces the deepest O2\mathrm{O}_2 features and a dense H2O\mathrm{H}_2\mathrm{O} line forest, whereas CH4\mathrm{CH}_4 is comparatively weak. The O2\mathrm{O}_2 A-band near 0.76 μm\mu\mathrm{m} provides the clearest visual tracer of the transition from anoxic to oxic conditions in the full optical-near-infrared spectrum.

Figure 7. Clear-atmosphere transmission spectra at R∼105R \sim10^{5} for representative Archean (A3-H1), Proterozoic (C1-V2), and modern Earth cases. Molecular contributions are shown over 0.4–1.8 μ\mum, with the lower panel emphasizing the growth of the O2_2 A-band as the atmosphere becomes oxygenated. This optical O2_2 feature lies outside the YJH wavelength range used for the ANDES detectability calculations.

The detailed Archean and Proterozoic spectral grids are shown in Appendix Figures 13 and 14. Across the Archean H1 families, increasing CH4_4/CO2_2 strengthens methane features throughout the bandpass. In the Proterozoic grid, two things are noticeable: increasing O2_2 introduces the A-band and other weaker O2_2 features, while increasing CH4_4 strengthens the methane bands.

These trends demonstrate that a single epoch label does not correspond to a unique spectrum; the observable state depends on the redox boundary conditions within that epoch.

Appendix Figure 15 provides a molecule-by-molecule decomposition of the modern spectrum. O2_2 and H2_2O dominate the strongest optical and near-infrared features in our wavelength range while CO2_2 and CH4_4 contribute weakly. This decomposition motivates the four molecules carried into the detectability analysis: O2_2 and CH4_4 as redox-sensitive tracers, and H2_2O and CO2_2 as environmental and climate context.

ANDES detectability

Figure 8 shows the observing requirements for reaching 3σ3\sigma and 5σ5\sigma detections for the representative solar-twin atmospheres. The methane-rich Archean model is most favorable for CH4_4, requiring 33 transits to reach 5σ5\sigma, while O2_2 remains effectively inaccessible because of the anoxic atmospheric composition. In the selected Proterozoic case, CH4_4 reaches 5σ5\sigma after 42 transits, whereas the O2_2 requirement exceeds 500 transits (200 equivalent observing nights) even at 3σ3\sigma. Modern Earth produces the strongest O2_2 signal of the three representative cases and reaches 3σ3\sigma after 38 transits, although a 5σ5\sigma detection still exceeds the required limit in the YJHonly configuration. It illustrates that atmospheric oxygenation strengthens the O2_2 signal substantially, but its detectability remains limited by the intrinsically small scale height of a high-mean-molecular-weight terrestrial atmosphere and by the exclusion of the strong 0.76 µm O2_2 A-band from the YJH-only configuration.

Figure 8. Estimated 5σ5\sigma and 3σ3\sigma detectability requirements in terms of transits (NtransitsN_{\mathrm{transits}}) and equivalent observing nights (NnightsN_{\mathrm{nights}}) for representative solar-host Archean, Proterozoic, and modern-Earth templates using ANDES YJH observations. Here, σ\sigma refers to the likelihood-based cross-correlation significance defined in § 2.3. The color scale shows log⁡10(Nnights)\log_{10}(N_{\mathrm{nights}}) for configurations within the observing limits. Cell annotations indicate NtransitsN_{\mathrm{transits}} and the corresponding number of equivalent 7 h observing nights. Configurations requiring more than 500 transits or 200 equivalent observing nights are shown as gray cells and marked n.d.

Figure 9 extends the spectral comparison across the full FGKM host-star grid. The heatmaps show the wavelength-integrated leave-one-out contribution of O2_2, CH4_4, H2_2O, and CO2_2 for each geological epoch. O2_2 contributes negligibly to the 3.9 Ga spectra for all hosts, as expected for the anoxic atmosphere, and its contribution increases toward the oxygenated epochs. CH4_4 shows a different pattern, with strong features persisting for several of the cooler stellar hosts even after atmospheric oxygenation. Hence, the molecular spectrum changes with both geological epoch and stellar host, reflecting the host-dependent atmospheric abundances produced by the photochemical models.

Figure 9. Feature-strength heatmaps for the FGKM hosts × epoch grid for O2_2, CH4_4, H2_2O, and CO2_2. Columns represent different hosts (hot F to cool M) and rows are geological epochs (3.9 Ga, 2.0 Ga, 0.8 Ga, modern). Color shows the wavelength-integrated positive leave-one-out transit-depth excess over 0.4–1.8 µm. CH4_4 dominates the integrated budget for cooler hosts, while O2_2 remains negligible at 3.9 Ga and strengthens mainly for late-M modern cases.

Figure 10 shows how these spectral differences translate into the number of transits and equivalent observing nights required for a 5σ detection with ANDES in the YJH bands. O2_2 does not reach a 5σ detection within the observing limits for the 3.9 Ga atmospheres and becomes gradually more detectable toward modern Earth. For modern Earth, the lowest equivalent observing-time requirements for O2_2 occur for the M3V and M8V proxies, requiring 9 and 20 transits, respectively, corresponding to approximately 5 equivalent observing nights in both cases. CH4_4 and CO2_2 are generally easier to detect around the cooler hosts; for the modern M3V case they require only 2 and 5 transits, respectively, compared with 20 transits for H2_2O. H2_2O also remains more demanding than CH4_4 and CO2_2 for the M8V proxy, requiring 51 transits in the modern case. The F0V and F7V cases remain outside the observing limits for all four molecules, while O2_2 and H2_2O also remain non-detectable for the G8V star.

Figure 10. 5σ requirement heatmaps for O₂, CH₄, H₂O, and CO₂ across FGKM hosts using ANDES YJH-band observations. Here, σ refers to the likelihood-based cross-correlation significance defined in § 2.3. The color scale shows log⁡10(Ntransits)\log_{10}(N_{\mathrm{transits}}) for configurations within the observing limits. Cell annotations give NtransitsN_{\mathrm{transits}} and the corresponding number of equivalent 7 h observing nights. Configurations requiring more than 500 transits or 200 equivalent observing nights are shown as gray cells and marked n.d.

Figure 11 provides one example of the cross-correlation signal underlying these detectability estimates. For the M8V host, the CH4_4 CCF changes substantially between geological epochs even though the observing setup remains unchanged. In the 225-frame example, the 3.9 Ga atmosphere remains below 3σ, whereas the 2.0 Ga, 0.8 Ga, and modern cases exceed 5σ. With the photon-noise scaling used for the detectability calculations, the 3.9 Ga case requires 28 transits to reach the 5σ criterion, compared with only 2–3 transits for the oxygenated epochs. This sequence follows the CH4_4 abundances obtained for the M8V photochemical models and illustrates why strong methane absorption is not restricted to the Archean atmosphere in the host-star grid.

Discussion

What is required to distinguish geological states?

The main result of our calculations indicates that geological state cannot be inferred from a single molecule. Methane provides the clearest example. The solar-host Archean grid is deliberately methane rich and produces strong CH4_4 signatures, whereas the 3.9 Ga models in the S. Rugheimer & L. Kaltenegger [64] grid use much lower CH4_4 boundary abundances. At the same time, cooler host stars can preserve substantial methane after oxygenation. A strong CH4_4 detection is compatible with both an anoxic, methane-rich atmosphere and an oxygenated atmosphere around a UV environment that permits CH4_4 to accumulate [71, 65].

O2_2 provides the principal discriminant between these cases, but only if the observations are sensitive to the relevant oxygenation level. In our Proterozoic grid, the pO2=10−7p\mathrm{O}_2 = 10^{-7} bar cases are spectroscopically much closer to the anoxic states than to the 10−310^{-3} bar cases. A non-detection of O2_2 thus constrains geological state only when the achieved upper limit lies below the O2_2 abundance expected for the intermediate atmosphere being tested. For this reason, reporting the sensitivity reached for O2_2 is as important as reporting a formal detection.

CO2_2 and H2_2O provide complementary environmental information. In the Earth-through-time grid, their abundances are tied to the implemented evolutionary boundary conditions, but on real planets neither molecule maps uniquely onto age. CO2_2 is governed by outgassing, weathering, and climate feedback, while H2_2O depends on the thermal structure and water inventory. Their role is primarily contextual, i.e., they help determine whether the O2_2-CH4_4 combination is physically plausible for the inferred climate state.

This distinction is also important for biosignature interpretation. An atmosphere classified as oxygenated is not automatically biological, and a geological-era analogue is not by itself a biosignature. Abiotic O2_2 production is especially relevant for M-dwarf planets ([47]). Any interpretation of joint O2_2 and CH4_4 therefore requires the broader atmospheric and stellar context, consistent with chemical-disequilibrium arguments ([40, 41]).

Relation to previous Earth-through-time and ANDES studies

Our atmospheric grid is directly motivated by K. Zahnle et al. [87], G. N. Arney et al. [1], and S. Rugheimer & L. Kaltenegger [64], who calculated self-consistent climate and photochemistry for Earth-like planets across four geological epochs and an FGKM stellar grid. Their study emphasised the low- to moderate-resolution spectral consequences of geological evolution. We retain the same conceptual structure but recompute the chemistry with Photochem, use modern MUSCLES UV spectra as host proxies, and propagate the resulting columns as inputs into R∼105R \sim10^{5} transmission spectra and ANDES cross-correlation studies. Differences in chemical network, fixed temperature structure, stellar UV inputs, and aerosol treatment prevent a one-to-one profile comparison.

The present study is complementary to the modern-Earth ANDES analysis of E. Kurzawa-Fernandez et al. (2026) in terms of detectability, but with the evolution of atmospheric chemistry more akin to M. H. Currie et al. [13] (but uses a larger grid of stellar proxies by including stars other than M dwarfs). E. Kurzawa-Fernandez et al. (2026) evaluated the detectabilities of H2_2O, CO2_2, CH4_4, and O2_2 for 18 known potentially habitable transiting planets, but by assuming the same modern-Earth atmosphere for all cases. Here, we instead vary both the geological state and host-star spectrum for idealized Earth twins. This means that our modern-Earth analogue is not similar to the modern-Earth modeled in E. Kurzawa-Fernandez et al. (2026), but instead a context dependent scenario based on both the geological epoch and the star. While both approaches identify cool M-dwarf systems as the most favorable transmission targets, our detectability time estimates for H2_2O are much higher compared to E. Kurzawa-Fernandez et al. (2026)’s estimates because CH4_4 starts dominating our modeled spectra for modern Earth atmospheres around such stars compared to always dominant H2_2O for their study. This discrepancy between H2_2O and CH4_4 is also seen in M. H. Currie et al. [13], who also use an epoch and host star dependent evolving atmospheric chemistry setup. They also find a trend of CH4_4 and CO2_2 being substantially easier to detect in comparison to O2_2 and H2_2O for modern Earth like analogues (labeled as pre-industrial Earth for their study) around M dwarf stars, akin to our study.

Our most favorable modern-Earth O2_2 cases require approximately 5 equivalent observing nights or 9 and 20 transits to reach a 5σ5\sigma detection threshold for the M3V and M8V hosts respectively, placing them towards the optimistic end of previous high-resolution studies. Earlier studies typically found requirements of tens to hundreds of transits for comparable O2_2 searches ([74, 60, 73, 13]) depending on the complexity of the simulations arising. ing from whether atmospheric refraction was included in spectrum generation or a moving exoplanet signal was injected or more realistic treatment of white and red noise was included, as well as the way a molecular detection was quantified (which differs across most papers). More recently, D. Deming et al. [17] showed that stellar heterogeneity can further complicate such observations: for a temperate rocky planet around an M4V star at 12 pc, their photon-noise-limited simulation reached only S/N≃4S/N \simeq4 for O2_2 after 200 transits, with unocculted star spots introducing an additional source of correlated noise. Direct comparison between these studies is, thus, difficult because the specific targets, wavelength ranges, noise prescriptions, and detection metrics are not uniform across them. In our case, the relatively short observing requirements are driven by the use of very nearby stellar proxies and by the assumption that the signal continues to improve according to photon-noise-limited N\sqrt{N} scaling after idealized removal of telluric and stellar contamination. Hence, the result-

ing observing times should be interpreted as optimistic estimates for the specific proxy systems considered here.

The YJH band restriction for ANDES as assumed here is especially relevant for O2_2. The 0.76 µm A-band, which dominates the optical comparison in Figure 7, is not included in the detectability maps. The YJH analysis presented here relies primarily on weaker near-infrared O2_2 features, including the 1.27 µm band, which is itself affected by terrestrial O2_2 and airglow. Extending the same framework to the optical ANDES arms will be essential for a complete assessment of oxygen detectability.

Host-star dependence and target selection

Our choice for cool hosts in the detectability maps is primarily motivated by two reasons. The first is geometric. For a fixed terrestrial atmosphere, the approximate amplitude of a transmission feature scales as RpH/R⋆2R_{\mathrm{p}}H/R_{\star}^{2}, making the same Earth twin substantially easier to observe around an M dwarf than around an F or G star. In our proxy grid, difference in contrast can vary over two orders of magnitude between the 1.57 R⊙R_{\odot} WASP-17 proxy and the 0.12 R⊙R_{\odot} TRAPPIST-1 proxy. Longer transits around larger stars do not compensate for that loss. The second contribution arises from the stellar radiation environment. The host UV spectrum controls photolysis and therefore the steady-state abundances of molecules such as CH₄, O₃, and N₂O. Consequently, a favorable stellar radius does not merely increase the transmission contrast; the stellar spectrum can also change the atmosphere being observed. This coupling is why a host-star grid is necessary when discussing geological analogues. The stellar proxies represent specific stars, and their distances affect the exposure times required to reach a given signal-to-noise ratio. Very nearby systems such as ε Eri and GJ 832 can saturate the CCDs even at the minimum exposure time, whereas more distant analogues would require longer integrations to achieve the same S/N per pixel. G. N. Arney et al. [1] spectroscopy. The reported Archean detectability estimates reflect as a cloud-free upper limit for our study. Second, the atmospheric temperature structures are assumed to be adiabatic profiles and are not iterated self-consistently with the atmospheric chemistry. The photochemical calculations are also one-dimensional and use constant vertical mixing and a single representative stellar zenith angle, thereby neglecting three-dimensional circulation and morning-evening terminator heterogeneity. Third, the list of opacities is incomplete in specific wavelength regions. Our O₃ opacity table begins at 1.4 μm and does not include the broad visible ozone absorption feature spanning approximately 0.45–0.75 μm [72]. Fourth, all planets are Earth twins with fixed mass, radius, and orbital impact parameter. Variations in surface gravity, scale height, cloud-top pressure, and atmospheric mixing can change the transmission signal substantially.

Model limitations and observational implications

Several assumptions make the present analysis optimistic. First, clouds and aerosol extinction are omitted from the transmission calculation. This is particularly important for the Archean models, for which organic haze is chemically active and would reduce the short-wavelength continuum available for transmission spectroscopy [1]. The reported Archean detectability estimates reflect as a cloud-free upper limit for our study. Second, the atmospheric temperature structures are assumed to be adiabatic profiles and are not iterated self-consistently with the atmospheric chemistry. The photochemical calculations are also one-dimensional and use constant vertical mixing and a single representative stellar zenith angle, thereby neglecting three-dimensional circulation and morning-evening terminator heterogeneity. Third, the list of opacities is incomplete in specific wavelength regions. Our O₃ opacity table begins at 1.4 μm and does not include the broad visible ozone absorption feature spanning approximately 0.45–0.75 μm [72]. Fourth, all planets are Earth twins with fixed mass, radius, and orbital impact parameter. Variations in surface gravity, scale height, cloud-top pressure, and atmospheric mixing can change the transmission signal substantially.

The detectability model introduces additional limitations. We assume photon-noise-limited N\sqrt{N} scaling from a reference stack, idealized treatment of telluric residuals, and a fixed duty-cycle prescription. Real H­RCCS observations are affected by wavelength-correlated telluric residuals, time-variable stellar spectra, stellar spots and faculae, imperfect detrending, and the Rossiter–McLaughlin effect [6 M. Brogi & M. R.] Line 2019; E. Palle et al. [54]. This is particularly relevant for O2_2, for which the planetary signal must be recovered in spectral regions containing strong terrestrial O2_2 absorption. Moreover, the present detectability calculations use only the YJH channels of ANDES; consequently, the O2_2 A-band at 0.76 μ\mum, as well as potentially useful optical CH4_4 and H2_2O features, do not contribute to the detectability maps.

Nearby bright proxies also approach the soft-saturation limits. In addition, the photon budgets are calculated using the actual distances of the stellar proxies. The absolute observing times correspond to these specific, in several cases exceptionally nearby, systems and should not be interpreted as representative of a volume-limited FGKM population. The resulting requirements should be interpreted primarily as a comparative measure of molecular detectability.

A further difference from M. H. Currie et al. [13] is the treatment of exposure time. They set each simulated exposure equal to the full transit duration to minimize the contribution of read noise, whereas our exposure times are determined separately for each stellar proxy to avoid detector saturation while maintaining a high S/N. As a result, a single transit can contain many individual exposures in our simulations, and the conversion from the required number of frames to the number of transits depends on the assumed cadence and readout overhead. Six of the eleven stellar proxies considered here have exposure times shorter than 2 s. Such short integrations may be difficult to realize in practice for full-frame observations. Differences in exposure-time prescription, together with the different noise treatments and detection metrics, also limit a direct comparison of our transit counts with those of M. H. Currie et al. [13].

Conclusions

We have investigated how the atmospheric evolution of an Earth-like planet would manifest in high-resolution transmission spectroscopy and how these signatures may be accessed with ELT/ANDES. We combine photochemical models spanning Archean, Proterozoic, and modern-Earth conditions with an FGKM host-star grid motivated by S. Rugheimer & L. Kaltenegger [64], propagate the resulting atmospheric compositions through petitRADTRANS, and evaluate their high-resolution detectability using the Ratri framework. Because O2_2 and CH4_4 provide the clearest contrast between reduced and oxidized atmospheric states in our models, Figure 12 summarizes how their joint detectability evolves across geological epochs and stellar hosts. Our main conclusions are summarized as follows:

Figure 12. Summary of 5σ5\sigma detection requirements with ELT/ANDES in the YJH band for O2_2 and CH4_4, following Figure 8 and Figure 10. The axes show the number of transits, NtrN_{\rm tr}, required for 5σ5\sigma detections of O2_2 and CH4_4, with detectability increasing toward the upper right corner. Each colored track follows one host star through the four geological epochs of the FGKM grid (circle: 3.9 Ga; square: 2.0 Ga; diamond: 0.8 Ga; star: modern), with arrows indicating the direction of evolution. Gray regions denote non-detections (Ntr>500N_{\rm tr} > 500 or more than 200 equivalent observing nights), while the shaded yellow region marks cases where both O2_2 and CH4_4 are detectable within 60 transits. Hosts without any 5σ5\sigma O2_2 or CH4_4 detection are not shown. For the cooler proxies, the tracks move from O2_2 non-detection at 3.9 Ga into the joint O2_2–CH4_4 detection region after oxygenation, whereas O2_2 remains outside the 5σ5\sigma observing limits for the solar-host cases.

  1. The dominant high-resolution molecular signatures evolve substantially with atmospheric redox state. The CH4_4-rich Archean models are characterized primarily by strong methane absorption, whereas O2_2 becomes increasingly important following atmospheric oxygenation and reaches its strongest spectral contribution under modern-Earth conditions. The Proterozoic calculations demonstrate that this transition is gradual: atmospheres with very low O2_2 abundances remain spectroscopically closer to anoxic states, while higher-O2_2 cases develop increasingly prominent oxygen features.

  1. CH4_4 abundance alone does not provide a unique indicator of an Archean-like atmosphere. Although methane is particularly prominent in the methane-rich Archean models, substantial CH4_4 can also persist in oxygenated atmospheres, especially around cooler host stars. The stellar UV

radiation field modifies the photochemical lifetime of methane and therefore changes the relationship between geological state and observable CH4_4 abundance. Consequently, a strong methane detection cannot by itself distinguish an anoxic early-Earth analogue from a later oxygenated atmosphere.

  1. The host star influences both the atmospheric chemistry and the observable transmission signal. The stellar spectral energy distribution controls photolysis and thus affects the abundances of species such as CH4_4, O3_3, and N2_2O, while the stellar radius directly determines the planet-to-star transmission contrast. These effects combine to produce substantial differences in molecular detectability across the FGKM host star grid. In our proxy sample, the shortest equivalent observing times generally occur for the M-dwarf proxies, although the nearby K2V case can also require relatively few transits for several molecules. The F-star cases are considerably more demanding because of their larger stellar radii and less favorable photon budgets.

  1. The relative detectability of individual molecules depends strongly on both geological epoch and stellar host. O2_2 does not reach the 5σ5\sigma criterion within the observing limits for any 3.9 Ga atmosphere, but becomes gradually more detectable after oxygenation. In the modern M3V, K2V, and M8V cases, O2_2 requires 9, 15, and 20 transits, respectively, while CH4_4 requires only 2, 4, and 3 and CO2_2 requires 5, 6, and 13. H2_2O is less accessible in the same cases, requiring 20, 12, and 51 transits. Thus, the most favorable oxygenated cool-star systems can provide access to both oxidized and reduced species within feasible observing campaigns under our assumptions.

  1. No single molecular detection uniquely identifies a geological state. The most informative diagnostic is the joint atmospheric context. In particular, the relative constraints on O2_2 and CH4_4 provide information on atmospheric redox state, while CO2_2 and H2_2O provide complementary constraints on the broader climatic and environmental conditions. A non-detection can also be informative only when the achieved abundance or spectral sensitivity is sufficiently constraining. Distinguishing atmospheric evolutionary states will need multi-molecule measurements.

  1. The absolute observing-time estimates should be interpreted as optimistic lower limits. The present calculations assume clear atmospheres, idealized telluric correction, photon-noise-limited N\sqrt{N} stacking, one-dimensional atmospheric structures, and specific nearby stellar proxies. The detectability calculations are additionally restricted to the ANDES YJH channels and therefore exclude the strong O2_2 A-band at 0.76 μ\mum. These assumptions affect the absolute number of transits or observing hours required for detection.

Acknowledgments

L.M. acknowledges funding support from the DAE through the NISER project RNI 4011.

Anthropic Claude Opus 5 was used to assist with figure generation, in accordance with the AAS guidelines on the use of large language models (https://journals.aas.org/author-llm-guidelines). All scientific code and analyses were developed by the authors, who verified all results and take full responsibility for the intellectual integrity of the work presented.

Author Contributions

D.D., S.D. and L.M. jointly developed the scientific framework and analysis plan. D.D. led the photochemical modeling, geological-grid analysis and transmission-spectrum calculations. Ratri ELT/ANDES simulation setup and subsequent high-resolution detectability analysis was performed by D.D. with inputs from S.D. The manuscript was written by D.D. and revised critically by S.D. and L.M. All authors contributed to the interpretation of the results and to the refinement of the final manuscript.

Software: astropy ([27, 2, 3]), numpy ([32]), scipy ([79]), astroplan ([51]), Photochem ([85]), petitRADTRANS ([49, 50]), Ratri ([15, 16]), ESO SkyCalc ([52, 35])

Appendix

Additional High-Resolution Transmission Spectra

This appendix presents the detailed high-resolution transmission spectra underlying the trends discussed in § 3. The figures provide a more complete view of how the molecular features change across the Archean and Proterozoic atmospheric grids and illustrate the individual molecular contributions to the modern-Earth spectrum.

Figure 13 shows the Archean spectral grid across the different climate families and CH4/CO2\mathrm{CH}_{4}/\mathrm{CO}_{2} boundary conditions. Figure 14 presents the corresponding Proterozoic grid, in which O2\mathrm{O}_{2} and CH4\mathrm{CH}_{4} are varied independently. Figure 15 provides a molecule-by-molecule decomposition of the modern-Earth spectrum and identifies the wavelength regions in which the principal absorbers contribute to the total transmission signal.

Figure 13. Archean spectral grid: rows are climate families A-D; the first three columns are rising surface CH4_4/CO2_2. Black curves show continuum-subtracted transit-depth excess; orange filled underlays show the CH4_4 contribution in the spectra. The rightmost column shows the spectral change within each family between the highest and lowest CH4_4/CO2_2 cases: continuum–subtracted excess of the highest CH4_4/CO2_2 case minus that of the lowest (teal curve; e.g. column 3 − column 1), with the orange underlay the corresponding high minus low CH4_4 contribution. Positive (negative) values indicate stronger (weaker) features at higher CH4_4/CO2_2; the difference panels use a zoomed vertical scale.

Figure 14. Proterozoic spectral grid: rows represent fixed surface pO2p\mathrm{O}_{2}; the first four columns are rising surface pCH4p\mathrm{CH}_{4}. Black curves show continuum-subtracted transit-depth excess; orange and purple filled underlays show the CH4\mathrm{CH}_{4} and O2\mathrm{O}_{2} contributions to the spectra, respectively. The rightmost column shows the spectral change within each O2\mathrm{O}_{2} row between the highest and lowest pCH4p\mathrm{CH}_{4} cases: continuum-subtracted excess of the highest-pCH4p\mathrm{CH}_{4} case minus that of the lowest (teal curve; e.g. column 4 − column 1), with orange and purple underlays the corresponding high minus low CH4\mathrm{CH}_{4} and O2\mathrm{O}_{2} contributions. Positive (negative) values indicate stronger (weaker) features at higher pCH4p\mathrm{CH}_{4}; the difference panels use a zoomed vertical scale.

Figure 15. Modern Earth clear-atmosphere transmission spectrum at R∼105R \sim10^{5} with molecule-by-molecule attribution over 0.40.4-1.8 μm1.8\,\mu\mathrm{m}. In all panels, species are color-coded as in the legend. (A) Continuum-subtracted transit-depth excess (Rp/Rs)2(R_{\mathrm{p}}/R_{\mathrm{s}})^{2} (black), measured relative to the 5th-percentile continuum floor (gray fill), with muted “only-molecule” traces of the active absorbers. (B) Molecular contributions Δ=(full−leave-one-out)\Delta= (\mathrm{full} - \mathrm{leave\text{-}one\text{-}out}) in ppm for the same wavelength; O2\mathrm{O}_{2} and H2O\mathrm{H}_{2}\mathrm{O} dominate selected optical-NIR intervals, while CO2\mathrm{CO}_{2}, CH4\mathrm{CH}_{4}, CO, HCN, C2H6\mathrm{C}_{2}\mathrm{H}_{6}, and O3\mathrm{O}_{3} provide weaker secondary features. (C) Offset contribution zooms over 0.680.68-1.30 μm1.30\,\mu\mathrm{m} isolating the O2\mathrm{O}_{2} B/A and 1.27 μm1.27\,\mu\mathrm{m} bands (upper) against overlapping H2O\mathrm{H}_{2}\mathrm{O} (lower). (D) Offset contribution zooms over 1.351.35-1.75 μm1.75\,\mu\mathrm{m} through the near-infrared H2O/CO2/CH4\mathrm{H}_{2}\mathrm{O}/\mathrm{CO}_{2}/\mathrm{CH}_{4} forest (top to bottom). O3\mathrm{O}_{3} contributions appear only where the O3\mathrm{O}_{3} line list is available (λ≳1.4 μm\lambda\gtrsim1.4\,\mu\mathrm{m}). Aerosols are not considered in the radiative transfer.

Host-Star Dependence of the Molecular Spectra

Figure 16-Figure 19 show the molecular contributions to the transmission spectra for the full stellar-proxy grid at each of the four geological epochs. These figures complement the host-star comparison presented in the main text by showing explicitly how the high-resolution molecular features change with stellar spectral type. The host-star dependence arises from both the different atmospheric compositions produced by the photochemical calculations and the changing stellar and planetary parameters used for the transmission spectra calculation. The figures are intended to illustrate the relative spectral behavior across the FGKM host star grid.

Figure 16. Molecule-isolated leave-one-out contributions for all eleven hosts over 0.4–1.8 µm for the anoxic 3.9 Ga epoch. Hosts are colored by TeffT_{\mathrm{eff}} (cool M crimson to hot F blue). Features are CH4\mathrm{CH}_{4}- and CO2\mathrm{CO}_{2}-dominated; O2\mathrm{O}_{2} remains negligible.

Figure 19. Same as Figure 16, but for modern Earth boundary conditions. O2_2 and H2_2O dominate selected intervals; CH4_4 is suppressed relative to earlier oxygenated epochs for many hosts.

Figure 17. Same as Figure 16, but for the 2.0 Ga (post-GOE) epoch. O2O_2 contributions begin to appear while CH4CH_4 remains strong, especially for cooler hosts.

Figure 18. Same as Figure 16, but for the 0.8 Ga (Neoproterozoic) epoch at higher oxygen.

ANDES Signal-to-Noise Ratios for the Stellar Proxy Grid

Figure 20 shows the wavelength-dependent continuum signal-to-noise ratio adopted in the Ratri simulations for each stellar proxy. The exposure times are adjusted to approach the assumed continuum count level while avoiding saturation where possible. This figure provides the instrumental context for the host-dependent detectability differences discussed in § 3. For the brightest nearby systems, the minimum exposure-time constraint limits this scaling. In particular, the K2V proxy ϵ\epsilon Eri remains above the nominal unsaturated count level at the 0.5 s exposure-time floor, while GJ 832 lies close to this regime.

References

  1. [1]Arney, G. N., Meadows, V. S., Domagal-Goldman, S. D., et al. 2017, ApJ, 836, 49, doi: 10.3847/1538-4357/836/1/49DOI
  2. [2]Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f
  3. [3]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74
  4. [4]Behr, P. R., France, K., Brown, A., et al. 2023, AJ, 166, 35, doi: 10.3847/1538-3881/acbd70
  5. [5]Bétrémieux, Y., & Kaltenegger, L. 2013, The Astrophysical Journal Letters, 772, L31
  6. [6]Birkby, J. L. 2018, in Handbook of Exoplanets, ed. H. J. Deeg & J. A. Belmonte, 16, doi: 10.1007/978-3-319-55333-7_16
  7. [7]Brogi, M., Giacobbe, P., Guilluy, G., et al. 2018, Astronomy & Astrophysics, 615, A16
  8. [8]Brogi, M., Kok, R. d., Albrecht, S., et al. 2016, The Astrophysical Journal, 817, 106
  9. [9]Brogi, M., & Line, M. R. 2019, The Astronomical Journal, 157, 114
  10. [10]Catling, D. C., & Zahnle, K. J. 2020, Science advances, 6, eaax1420
  11. [11]Cheverall, C. J., & Madhusudhan, N. 2024, The Astronomical Journal, 167, 272
  12. [12]Claire, M. W., Sheets, J., Cohen, M., et al. 2012, ApJ, 757, 95, doi: 10.1088/0004-637X/757/1/95DOI
  13. [13]Currie, M. H., Meadows, V. S., & Rasmussen, K. C. 2023, The Planetary Science Journal, 4, 83DOI
  14. [14]Dash, S., Brogi, M., Gandhi, S., et al. 2024, Monthly Notices of the Royal Astronomical Society, 530, 3100
  15. [15]Dash, S., Brogi, M., Seidler, F. L., et al. 2025, MNRAS, 538, 3042, https://arxiv.org/abs/2503.15196
  16. [16]Dash, S., Dubey, D., & Majumdar, L. 2026, AJ, 172, 90, https://arxiv.org/abs/2602.22830
  17. [17]Deming, D., Currie, M. H., Meadows, V. S., & Peacock, S. 2025, The Astronomical Journal, 170, 11
  18. [18]Dodd, M. S., Papineau, D., Grenne, T., et al. 2017, Nature, 543, 60
  19. [19]Driese, S. G., Jirsa, M. A., Ren, M., et al. 2011, Precambrian Research, 189, 1
  20. [20]Dubey, D., & Majumdar, L. 2024, The Astrophysical Journal, 972, 165
  21. [21]Dubey, D., Majumdar, L., Beichman, C., et al. 2025, The Astrophysical Journal Supplement Series, 278, 19
  22. [22]France, K., Loyd, R. O. P., Youngblood, A., et al. 2016, ApJ, 820, 89, doi: 10.3847/0004-637X/820/2/89DOI
  23. [23]Gandhi, S., Brogi, M., & Webb, R. K. 2020, Monthly Notices of the Royal Astronomical Society, 498, 194
  24. [24]Giacobbe, P., Brogi, M., Gandhi, S., et al. 2021, Nature, 592, 205
  25. [25]Gordon, I. E., Rothman, L. S., Hargreaves, R. J., et al. 2022, Journal of quantitative spectroscopy & radiative transfer, 277, 1
  26. [26]Gordon, I. E., Rothman, L., Hargreaves, R. J., et al. 2026, Journal of Quantitative Spectroscopy and Radiative Transfer, 353, 109807
  27. [27]Greenfield, P., Robitaille, T., Tollérud, E., et al. 2013, http://ascl.net/1304.002
  28. [28]Grimm, S. L., Malik, M., Kitzmann, D., et al. 2021, The Astrophysical Journal Supplement Series, 253, 30
  29. [29]Hardegree-Ullman, K. K., Apai, D., Bergsten, G. J., Pascucci, I., & López-Morales, M. 2023, The Astronomical Journal, 165, 267
  30. [30]Hargreaves, R. J., Gordon, I. E., Rey, M., et al. 2020, ApJS, 247, 55, doi: 10.3847/1538-4365/ab7a1aDOI
  31. [31]Hargreaves, R. J., Gordon, I. E., Rothman, L. S., et al. 2019, Journal of Quantitative Spectroscopy and Radiative Transfer, 232, 35
  32. [32]Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2DOI
  33. [33]Harris, G. J., Polyansky, O. L., & Tennyson, J. 2002, The Astrophysical Journal, 578, 657
  34. [34]Holland, H. D. 2006, Philosophical Transactions of the Royal Society B: Biological Sciences, 361, 903
  35. [35]Jones, A., Noll, S., Kausch, W., Szyszka, C., & Kimeswenger, S. 2013, Astronomy & Astrophysics, 560, A91
  36. [38]Kaltenegger, L., Traub, W. A., & Jucks, K. W. 2007, The Astrophysical Journal, 658, 598
  37. [39]Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131, doi: 10.1088/0004-637X/765/2/131DOI
  38. [40]Krissansen-Totton, J., Olson, S., & Catling, D. C. 2018, Science Advances, 4, eaao5747, doi: 10.1126/sciadv.aao5747
  39. [41]Krissansen-Totton, J., Thompson, M., Galloway, M. L., & Fortney, J. J. 2022, Nature Astronomy, 6, 189, doi: 10.1038/s41550-021-01579-7
  40. [42]Kurzawa-Ferrandez, E., Bello-Arufe, A., & Hu, R. 2026, The Astronomical Journal, 172, 165
  41. [43]Lederberg, J. 1965, Nature, 207, 9
  42. [44]Lovelock, J. E. 1965, Nature, 207, 568
  43. [45]Lyons, T. W., Reinhard, C. T., & Planavsky, N. J. 2014, Nature, 506, 307
  44. [46]McLaughlin, D. 1924, Astrophysical Journal, 60, 22-31 (1924), 60
  45. [47]Meadows, V. S., Reinhard, C. T., Arney, G. N., et al. 2018, Astrobiology, 18, 630
  46. [48]Mojzsis, S. J., Arrhenius, G., McKeegan, K., et al. 1996, Nature, 384, 55
  47. [49]Mollière, P., Wardenier, J. P., van Boekel, R., et al. 2019, A&A, 627, A67, doi: 10.1051/0004-6361/201935470DOI
  48. [50]Mollière, P., Stölker, T., Lacour, S., et al. 2020, A&A, 640, A131, doi: 10.1051/0004-6361/202038325DOI
  49. [51]Morris, B. M., Tollerud, E., Sipőcz, B., et al. 2018, The Astronomical Journal, 155, 128, doi: 10.3847/1538-3881/aaa47eDOI
  50. [52]Noll, S., Kausch, W., Barden, M., et al. 2012, Astronomy & Astrophysics, 543, A92
  51. [53]Olson, S. L., Schwieterman, E. W., Reinhard, C. T., & Lyons, T. W. 2018, arXiv preprint arXiv:1803.05967DOI
  52. [54]Palle, E., Biazzo, K., Bolmont, E., et al. 2025, Experimental Astronomy, 59, 29
  53. [56]Quirrenbach, A., Amado, P. J., Mandel, H., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7735, Ground-based and Airborne Instrumentation for Astronomy III, ed. I. S. McLean, S. K. Ramsay, & H. Takami, 773513, doi: 10.1117/12.857777DOI
  54. [57]Ribas, I., Guinan, E. F., Güdel, M., & Audard, M. 2005, ApJ, 622, 680, doi: 10.1086/429777DOI
  55. [58]Robinson, T. D., Fortney, J. J., & Hubbard, W. B. 2017, The Astrophysical Journal, 850, 128
  56. [59]Robinson, T. D., Meadows, V. S., Crisp, D., et al. 2011, Astrobiology, 11, 393
  57. [60]Rodler, F., & López-Morales, M. 2014, The Astrophysical Journal, 781, 54
  58. [61]Rossiter, R. 1924, Astrophysical Journal, 60, 15-21 (1924), 60
  59. [62]Rothman, L. S., Gordon, I. E., Barber, R. J., et al. 2010, JQSRT, 111, 2139, doi: 10.1016/j.jqsrt.2010.05.001DOI
  60. [63]Rothman, L. S., Gordon, I. E., Babikov, Y., et al. 2013, Journal of Quantitative Spectroscopy and Radiative Transfer, 130, 4
  61. [64]Rugheimer, S., & Kaltenegger, L. 2018, The Astrophysical Journal, 854, 19
  62. [65]Rugheimer, S., Kaltenegger, L., Segura, A., Linsky, J., & Mohanty, S. 2015, The Astrophysical Journal, 809, 57
  63. [66]Rugheimer, S., Kaltenegger, L., Zsom, A., Segura, A., & Sasselov, D. 2013, Astrobiology, 13, 251
  64. [67]Sagan, C., Thompson, W. R., Carlson, R., Gurnett, D., & Hord, C. 1993, Nature, 365, 715
  65. [68]Sánchez-López, A., Alonso-Floriano, F. J., López-Puertas, M., et al. 2019, Astronomy & Astrophysics, 630, A53
  66. [69]Schwieterman, E. W., Kiang, N. Y., Parenteau, M. N., et al. 2018, Astrobiology, 18, 663
  67. [70]Seager, S., Schrenk, M., & Bains, W. 2012, Astrobiology, 12, 61
  68. [71]Segura, A., Kasting, J. F., Meadows, V., et al. 2005, Astrobiology, 5, 706
  69. [72]Serdyuchenko, A., Gorshelev, V., Weber, M., Chehade, W., & Burrows, J. P. 2014, Atmospheric Measurement Techniques, 7, 625, doi: 10.5194/amt-7-625-2014DOI
  70. [73]Serindag, D. B., & Snellen, I. A. 2019, The Astrophysical Journal Letters, 871, L7
  71. [74]Snellen, I., De Kok, R., Le Poole, R., Brogi, M., & Birkby, J. 2013, The Astrophysical Journal, 764, 182
  72. [75]Snellen, I., de Kok, R., Birkby, J., et al. 2015, Astronomy & Astrophysics, 576, A59
  73. [76]Stassun, K. G., Collins, K. A., & Gaudi, B. S. 2017, AJ, 153, 136, doi: 10.3847/1538-3881/aa5df3DOI
  74. [77]Thompson, M. A., Krissansen-Totton, J., Wogan, N., Telus, M., & Fortney, J. J. 2022, Proceedings of the National Academy of Sciences, 119, e2117933119
  75. [78]Trainer, M. G., Pavlov, A. A., DeWitt, H. L., et al. 2006, Proceedings of the National Academy of Sciences, 103, 18035, doi: 10.1073/pnas.0608561103DOI
  76. [79]Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, 17, 261, doi: 10.1038/s41592-019-0686-2DOI
  77. [80]Western, C. M., Carter-Blatchford, L., Crozet, P., et al. 2018, Journal of Quantitative Spectroscopy and Radiative Transfer, 219, 127
  78. [81]Wilks, S. S. 1938, The annals of mathematical statistics, 9, 60
  79. [82]Wilson, D. J., Froning, C. S., Duvvuri, G. M., et al. 2021, ApJ, 911, 18, doi: 10.3847/1538-4357/abe771DOI
  80. [83]Wilson, D. J., Froning, C. S., Duvvuri, G. M., et al. 2025, ApJ, 978, 85, doi: 10.3847/1538-4357/ad9251DOI
  81. [84]Winn, J. N. 2010, in Exoplanets, ed. S. Seager, 55–77, doi: 10.48550/arXiv.1001.2010DOI
  82. [85]Wogan, N. F., Batalha, N. E., Zahnle, K., et al. 2025, PSJ, 6, 256, doi: 10.3847/PSJ/ae0e1cDOI
  83. [86]Wyttenbach, A., Ehrenreich, D., Lovis, C., Udry, S., & Pepe, F. 2015, Astronomy & Astrophysics, 577, A62
  84. [87]Zahnle, K., Claire, M., & Catling, D. 2006, Geobiology, 4, 271, doi: 10.1111/j.1472-4669.2006.00085.xDOI
  85. [88]Zandian, M., Farris, M., McLeVige, W., et al. 2016, in High Energy, Optical, and Infrared Detectors for Astronomy VII, Vol. 9915, SPIE, 148–158

Paper details

vlnc:W_8t4zuquf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Claude Opus 5 (bare)

View statement
Looking for related papers…
Contents