Timing and Likelihood of the Origin of Life Derived from Post-Impact Highly Reducing Atmospheres

Abstract

Big impacts on the early Earth would have created highly reducing atmospheres that generated molecules needed for the origin of life, such as nitriles. However, such impactors could have been followed by collisions that were sufficiently big to vaporize the ocean and destroy any pre-existing life. Thus, a post-impact-reducing atmosphere that gives rise to life needs to be followed by a lack of subsequent sterilizing impacts for life to persist. We assume that prebiotic chemistry required a post-impact-reducing atmosphere. Then, using statistics for the impact history on Earth and the minimum impact mass needed to generate post-impact highly reducing atmospheres, we show that the median timing of impact-driven biopoiesis is favored early in the Hadean, ∼4.35 Ga. However, uncertainties are large because impact bombardment is stochastic, and so biopoiesis could have occurred between 4.45 and 3.9 Ga within 95% uncertainty. In an optimistic scenario for biopoiesis from post-impact-reducing atmospheres, we find that the origin of life is favorable in ∼90% of stochastic impact realizations. In our most pessimistic case, biopoiesis is still fairly likely (∼20% chance). This potentially bodes well for life on rocky exoplanets that have experienced an early episode of impact bombardment given how planets form.

1. Introduction

Benner et al. (2020) argued that the most likely time for the origin of life (OoL) would have been 4.36 ± 0.1 Ga in the wake of a hypothesized $\sim 2\times {10}^{22}$ kg (∼2100 km diameter) asteroid impact called Moneta. These authors suggested that iron delivered by Moneta’s core would have reacted with impact-vaporized ocean water to generate a reducing Hadean atmosphere conducive to the photochemical generation of essential prebiotic nitriles like hydrogen cyanide (HCN) and cyanoacetylene (HCCCN), although their production rates were not modeled. Nitriles are required in prebiotic schemes that synthesize ribonucleotide precursors to RNA (Patel et al., 2015; Powner et al., 2009; Becker et al., 2019), and these molecules can demonstrably be made homochiral (Ozturk et al., 2023a), which, in turn, can propagate homochirality throughout biochemistry to peptides and metabolites (Ozturk et al., 2023b). A transient period of profuse nitrile production after a Moneta impact could have been a window of opportunity for the origin of RNA and, ultimately, life.

A single Moneta impact is one hypothesis for the relative excess of highly siderophile elements (i.e., iron-loving, abbreviated HSEs) in the Moon’s and Earth’s mantles. During the Moon-forming impact, HSEs should have been sequestered in each planetary core, so over-abundant HSEs in today’s mantles are commonly interpreted as evidence for late-accretion impactors. Earth’s mantle has substantially more HSEs than the Moon by an amount that cannot be accounted for by Earth’s greater gravitational cross section (Day and Walker, 2015). A single massive Moneta impact could explain the Earth-Moon HSE discrepancy because, by the statistics of small numbers, Moneta could have missed the Moon and hit the Earth (Sleep et al., 1989; Bottke et al., 2010).

However, the lunar HSE depletion can be explained without a Moneta impact. Lunar HSEs could have been lost to space during impact delivery because of the Moon’s small gravity (Kraus et al., 2015). Alternatively, HSEs delivered to the Moon during its ∼150 million-year magma ocean could have been sequestered in the core due to iron sulfide exsolution (Morbidelli et al., 2018; Rubie et al., 2016). Finally, there is some chance that Earth’s HSEs do not record late impacts because the Moon-forming impact-delivered HSEs (Sleep, 2016). Another explanation for Earth’s HSEs that do not require impacts is that HSEs are gradual core contributions over time from mantle plumes (Halliday and Canup, 2023; Mundl-Petermeier et al., 2020). If indeed Earth’s HSEs reflect asteroid bombardment after the Moon formed, then the HSEs can be explained by multiple ∼500–2000 km impacts rather than a single big (∼2100 km) collision.

Recently, Wogan et al. (2023) and Zahnle et al. (2020) used photochemical models of post-impact atmospheres to show that impacts significantly smaller than Moneta can efficiently produce molecules critical for prebiotic synthesis. In their simulations, impactor iron equilibrates with vaporized ocean water to generate atmospheric H₂, CH₄, and NH₃ that form thermochemically as the reducing steam atmosphere cools. Once steam condenses to an ocean, subsequent photochemistry of a Titan-like, hazy atmosphere produces prebiotic nitriles.

Zahnle et al. (2020) contributed a preliminary study that used simple atmospheric models and assumed all impact-delivered iron was available to reduce the atmosphere, and later Wogan et al. (2023) improved upon their calculations with more sophisticated and accurate simulations. Photochemical modeling in Wogan et al. (2023) showed that the production of both HCN and HCCCN only occurs when ${\text{CH}}_4/{\text{CO}}_2\mathrm{\gtrsim }0.1$ , while the atmosphere is hazy. In an optimistic scenario, which includes nickel-catalyzed methane production, results suggest that ${\text{CH}}_4/{\text{CO}}_2>0.1$ (i.e., significant nitriles) occurs for impacts $>4\times {10}^{20}$ kg (>570 km diameter). In a pessimistic case, which assumes that only a fraction of impactor iron reacts with the atmosphere (Citron and Stewart, 2022; Marchi et al., 2018), a $>5\times {10}^{21}$ kg (>1330 km diameter) impactor is required to produce substantial HCN and HCCCN (Wogan et al., 2023). Both size limits may even be conservative, given experimental results that show additional catalysts can generate methane (Hill and Nuth, 2003; Nuth et al., 2016, 2020). Regardless of uncertainty, atmospheric models suggest that an impactor much smaller than Moneta could deliver the nitriles needed for an RNA OoL.

Here, we employ the results of Wogan et al. (2023) along with Monte-Carlo simulations of Earth’s impact history to estimate when life most likely emerged, assuming the OoL required abundant prebiotic nitriles (e.g., HCN). Our calculations account for the possibility of planet sterilization by impacts that would have vaporized the ocean (Sleep et al., 1989; Maher and Stevenson, 1988). We assume that an OoL-favorable impact is one that produces significant prebiotic nitriles and is not subsequently followed by an ocean-vaporizing impact that destroys any biosphere that has emerged. By considering the fraction of stochastic impact realizations that do not have an OoL-favorable impact, we also estimate the probability of life beginning if Earth’s history was rerun.

2. Methods

The rate impacts that hit Earth bigger than mass m (in kg) at age t (in Gyr) can be written as follows:$$f(t,m)=F_0(t)S_0(m)$$(1)

Here, $S_0(m)$ is the size-frequency distribution of impactors normalized to a reference mass m ₀ so that $S_0(m_0)=1$ . Also, $F_0(t)$ is the number of impacts per billion years with mass greater than m ₀.

We assume that the majority of the size-frequency distribution of impactors is identical to that of the main-belt asteroids (Extended Data Figure 1 in Marchi et al., 2014). Data for the frequency of main-belt asteroids only extends down to about 1 km diameter ( $1.3\times {10}^{12}$ kg). To extend the distribution to smaller objects, we use the observed 1400 ratio between the frequency of >1 and >20 km craters on the Moon following the work of Morbidelli et al. (2018). The Johnson et al. (2016) crater-scaling relation suggests that 1 and 20 km craters on the Moon correspond to 41 and 1220 m asteroids, respectively. In this crater-scaling calculation, we use an average lunar impact velocity of 13 km/s as appropriate for the leftover planetesimals that would have been primarily responsible for the >1 km craters on the Moon in our assumed bombardment flux (Fig. 1b; Marchi, 2021; Morbidelli et al., 2018). The largest main-belt asteroid is ∼1000 km, so we must extrapolate to larger impactors. Through the main text, we use the same extrapolation as that of Marchi et al. (2014), which extends the distribution to big impacts with a slope $d(\ln \hspace{0.17em}{S}_0)/d(\ln \hspace{0.17em}m)=-0.415$ (Fig. 1). In Appendix A, we show that alternatively extrapolating with the red dashed line in Figure 1a (with slope $d(\ln \hspace{0.17em}{S}_0)/d(\ln \hspace{0.17em}m)=-1.0$ ) does not significantly change our overall conclusions. Finally, we normalize the size-frequency distribution to the impact mass required to make a 1 km crater on the Moon (41 m object, $m_0=9\times {10}^7$ kg). Figure 1a shows the resulting size-frequency distribution.

FIG. 1.

Our assumed (a) size-frequency distribution (SFD) of impactors and (b) lunar cratering record. Most of the size-frequency distribution is identical to the main-belt asteroids [Extended Data Figure 1 in Marchi et al., 2014]. We extrapolate to objects $\lesssim {10}^{11}$ kg using the observed frequency ratio of >1 and >20 km lunar craters following Morbidelli et al. [2018] and Marchi [2021], and, in the main text, log-log extrapolate to asteroids $\mathrm{\gtrsim }{10}^{21}$ kg following Marchi et al. [2014]. In Appendix A, we also consider the red dashed extrapolation to $\mathrm{\gtrsim }{10}^{21}$ kg impacts. The lunar cratering record is the red line in Figure 5 of Morbidelli et al. [2018].

The flux of impactors, $F_0(t)$ , can be estimated from the lunar cratering record. We adopt the accretionary tail scenario discussed by Morbidelli et al. (2018). In other words, we assume that the “late-heavy bombardment” did not happen (Cartwright et al., 2022; Hartmann, 2019; Zellner, 2017) and that the impact flux on Earth was monotonically decreasing throughout the Hadean eon. Figure 1b shows n_m , our assumed lunar impact history taken from the work of Morbidelli et al. (2018) (red line in their Fig. 5). We use the parameterization $n_m={10}^{a{e}^{bt}+c}$ , where a, b, and c are parameter fits given in Figure 1b.

Extrapolating the lunar impact history to Earth requires a correction for Earth’s greater gravitational attraction. Assuming that the approach velocity of impactors far from Earth and the Moon was on average ∼13 km/s (Marchi et al., 2018) and using Equation (7)–(14) from the work of Zahnle and Sleep (2006), we compute that Earth should have received $s_0=1.68$ times more impacts per unit surface area than the Moon. Therefore, the number of impacts on Earth should be$$N_0=A_{\oplus }{s}_0{n}_m=A_{\oplus }{s}_0{10}^{a{e}^{bt}+c}$$(2)

Equation (2) also accounts for Earth’s surface area ( $A_{\oplus }$ ), which makes N ₀ the number of impacts on Earth since age t that would cause a crater >1 km on the Moon. As discussed previously, the Johnson et al. (2016) crater-scaling relation finds that a 1 km lunar crater corresponds to a 41 m object with mass $m_0=9\times {10}^7$ kg assuming a 13 km/s lunar impact velocity.

The time derivative of N ₀ gives the flux, F ₀:$$F_0(t)=\frac{d{N}_0}{dt}$$(3)

The average number of impacts on Earth between time t ₁ and t ₂ with mass greater than m is then$$\begin{array}{ll}N(m,t_1,t_2)\hfill & ={\displaystyle {\int }_{t_1}^{t_2}f}(t,m)dt\hfill \\ \hfill & =S_0(m){\displaystyle {\int }_{t_1}^{t_2}\frac{d{N}_0}{dt}}dt\hfill \\ \hfill & =S_0(m)A_{\oplus }{s}_0\left({10}^{a{e}^{b{t}_2}+c}-{10}^{a{e}^{b{t}_1}+c}\right)\hfill \end{array}$$(4)

To simulate impact histories, we consider a grid of ∼200 impactor masses between 10¹² and 10²³ kg, and a grid of ∼60 times between 4.5 and 3.5 Ga. Indexes j and i indicate the mass and time grid cells, respectively, while, for example, $j-\frac{1}{2}$ and $j+\frac{1}{2}$ indicate the edges of the grid cell. We can compute the expected number of impacts within a mass and time grid cell with the following:$${\overline{N}}_{ij}=N(m_{j-\frac{1}{2}},t_{i-\frac{1}{2}},t_{i+\frac{1}{2}})-N(m_{j+\frac{1}{2}},t_{i-\frac{1}{2}},t_{i+\frac{1}{2}})$$(5)

Next, we sample a Poisson distribution for each ${\overline{N}}_{ij}$ , giving a stochastic number of impacts in each mass and time grid cell, which constitutes an impact history. By modeling impacts as a Poisson process, we are assuming that each impact does not change the probability of the next one (i.e., impacts are independent). Performing this sampling thousands of times captures many of the possible impact histories. We only keep impact histories that accrete a total mass of $2\times {10}^{22}$ to $6\times {10}^{22}$ kg because this is approximately the mass implied by the HSEs in Earth’s mantle, assuming that HSEs were delivered by late accretion. This range of masses is based on the work of Day and Walker (2015), who used mantle HSEs to suggest that Earth was impacted by $\sim 3\times {10}^{22}$ to $4.8\times {10}^{22}$ kg of material, but we choose a wider range of masses because HSEs could have been lost to Earth’s core or to space during impacts (Marchi et al., 2018).

Our final step is to assign impact velocities to each collision in the many sampled impact histories. Appendix Figure A1 shows Earth’s impact velocity distribution. We created this distribution by using the JPL database of close approaches to Earth by small bodies (Park and Chamberlin, 2023), considering all close-approach asteroids within 0.05 AU to Earth. The database gives the approach velocity of each asteroid far from Earth, so we compute the impact velocity by accounting for Earth’s gravitational potential energy. Here, we are assuming that the impact velocity distribution for small bodies (e.g., < 10 km) is identical to the velocity distribution for bigger asteroids (e.g., > 10 km). This may not be the case because small bodies are more strongly influenced by processes like the Yarkovsky effect (Bottke et al., 2006), where an asteroid’s orbit is altered by anisotropic thermal emission.

The final collection of sampled impact histories can then be used to compute impact statistics that are relevant to the timing and likelihood of the OoL.

3. Results

Figure 2 illustrates three of 5000 impact histories computed with our Monte-Carlo approach (Section 2). In Figure 2a, a Moneta-sized impact, which we define as an impact between $2\times {10}^{22}$ and $6\times {10}^{22}$ kg, occurs at ∼4.46 Ga delivering nearly all of Earth’s mantle HSEs. The impact should also chemically reduce the Hadean atmosphere creating conditions favorable for prebiotic chemistry. However, in this case, Moneta cannot have given rise to extant life on Earth because Moneta is followed by an ocean-vaporizing collision at ∼4.35 Ga, which should have destroyed any existing biosphere. Based on smoothed-particle hydrodynamics impact simulations, we nominally assume that 10% of an impact’s kinetic energy goes into vaporizing the ocean (Appendix B, Citron and Stewart, 2022). Vaporizing the ocean requires $5\times {10}^{27}$ J, so a $>5\times {10}^{28}$ J impact should deliver sufficient energy under our assumptions. Appendix B considers alternative criteria for ocean vaporization. In Figure 2a, the last ocean-vaporizing impact is only $\sim 1.6\times {10}^{20}$ kg (420 km) with an impact velocity of 30.3 km/s, giving it about $7.3\times {10}^{28}$ J of kinetic energy.

FIG. 2.

Three simulated impact histories of the 5000 derived from our Monte-Carlo approach (Section 2). Each vertical line indicates an impact of a mass shown by the y-axis. The red lines indicate the last impact to vaporize the ocean assuming 10% of an impactor’s kinetic energy heats the ocean. The legend in (a) gives conversions between impact mass and diameter for 500, 1000, and 2000 km asteroids.

Moneta also occurs in the Figure 2b impact realization. In this scenario, any life created in the wake of Moneta will not be subsequently impact-exterminated because Moneta is the last impact to vaporize the ocean.

Figure 2c shows an impact history where Moneta does not occur. Instead, Earth’s mantle HSEs are delivered by multiple 10²¹ to $4\times {10}^{21}$ kg collisions. Whether this bombardment history would favor biopoiesis depends on the required minimum impact mass to produce significant OoL nitriles (e.g., HCN and HCCCN). Wogan et al. (2023) used photochemical models of post-impact atmospheres to show that, optimistically, an impact $>4\times {10}^{20}$ kg should give rise to an atmosphere that makes significant HCN and HCCCN. Under this optimistic requirement, the last ocean-vaporizing impact in Figure 2c would be favorable for life’s origin and occurs at ∼4.25 Ga with mass $5\times {10}^{20}$ kg. However, with pessimistic modeling assumptions, Wogan et al. (2023) found that a $>5\times {10}^{21}$ kg impactor is instead required for substantial nitrile delivery to the surface. In this alternative scenario, an environment conducive to an OoL never occurs in Figure 2c because there is no impactor larger than $5\times {10}^{21}$ kg.

As illustrated in Figure 2, an OoL-favorable impact does not occur in every simulated impact history. In our impact-driven model for the OoL, biopoiesis requires (1) an impact of sufficient mass to make prebiotic nitriles and (2) the lack of a later ocean-vaporizing collision that destroys the biosphere without rekindling it. Figure 3 shows the probability of both conditions occurring as a function of the minimum impact mass that produces prebiotic molecules.

FIG. 3.

The probability of an impact that causes favorable conditions for an origin of life on Earth. The blue solid line is the probability of an impact occurring at least once. The orange dashed line is the probability of an impact without a subsequent ocean-vaporizing collision. Probabilities are shown as a function of the minimum impact mass to produce significant prebiotic molecules (e.g., HCN). Two plausible minimum masses are the Wogan et al. [2023] optimistic and pessimistic scenarios indicated with vertical dotted lines. The shaded region between $2\times {10}^{22}$ and $6\times {10}^{22}$ kg are Moneta-sized impacts because they deliver almost all of Earth’s mantle HSEs. There is a 93% chance of an impact creating favorable origin of life conditions in the Wogan et al. [2023] optimistic case and a 30% chance in the pessimistic case. The chance of a Moneta-size impact giving rise to life that is not subsequently destroyed is only 6%.

For example, consider the Wogan et al. (2023) optimistic minimum mass of $4\times {10}^{20}$ kg indicated in Figure 3. All simulated impact histories have at least one impact bigger than $4\times {10}^{20}$ kg (blue line in Fig. 3 is 1.0 at the given mass). However, the probability of an OoL-favorable impact in this scenario is 93%, because 7% of the time the last $4\times {10}^{20}$ kg collision is followed by a smaller ocean-vaporizing impact that perhaps sterilizes the planet (orange dashed line in Fig. 3 with 0.93 at $4\times {10}^{20}$ kg). An OoL is less probable if we consider the Wogan et al. (2023) pessimistic minimum mass ( $>5\times {10}^{21}$ kg) to produce significant OoL nitriles. A $>5\times {10}^{21}$ kg impact without later ocean vaporization occurs 30% of the time. An OoL-favorable impact was, therefore, relatively likely on the early Earth and strongly favored (93% probability) under optimistic assumptions.

To estimate the most likely timing for the OoL, we consider both the optimistic and pessimistic criteria of Wogan et al. (2023). Figure 4a is the optimistic case, showing the timing (Fig. 4a (i)) and mass (Fig. 4a (ii)) of the last $>4\times {10}^{20}$ kg impactor for the 93% of impact histories that do not experience later ocean vaporization. The median timing of an OoL-favorable impact is 4.31 Ga with 95% uncertainty between 3.86 and 4.44 Ga. The large uncertainty mostly stems from the stochastic, Poisson nature of asteroid bombardment. The probability distribution for the mass of the OoL-favorable impact peaks at $4\times {10}^{20}$ kg and decreases with increasing mass because of the size-frequency distribution of asteroids (Fig. 1a). There is only a 4% chance that the impactor is Moneta-sized (i.e., between $2\times {10}^{22}$ and $6\times {10}^{22}$ kg).

FIG. 4.

The timing and mass of an OoL-favorable impact on the early Earth. (a) considers an optimistic minimum impact mass needed to generate the molecules for an RNA origin of life and (b) examines a pessimistic minimum impact mass [Wogan et al., 2023]. In either (a) or (b), panels (i) and (ii) show probability distributions for the timing and mass, respectively, of the last impact that produces prebiotic molecules which do not have a later ocean-vaporizing collision. The vertical dashed lines on (a) i and (b) i plot are median values, and the vertical dotted lines are the 95% confidence intervals. In an impact-driven origin of life, biopoiesis most likely occurred at 4.31 Ga (95% CI: 3.86 to 4.44 Ga) assuming panel (a) and at 4.34 Ga (95% CI: 3.86 to 4.46 Ga) assuming panel (b).

The timing of an OoL impact is qualitatively unchanged when instead adopting the Wogan et al. (2023) pessimistic case, which requires a $>5\times {10}^{21}$ kg impact without later ocean vaporization (Fig. 4b). In this scenario, the median timing for life’s origin is 4.34 Ga with a 95% confidence interval spanning 3.86–4.46 Ga (Fig. 4b (i)). Under these assumptions, the OoL-favorable impact is Moneta-sized 20% of the time with no subsequent sterilization (Fig. 4b (ii)), which is more probable compared to the “optimistic” case (Fig. 4a (ii)).

In this model for biopoiesis, we suggest that life could have started within ∼10 s of millions of years after an OoL-favorable impact because this is the maximum duration of the H₂- and CH₄-rich atmosphere that makes nitriles like HCN (Wogan et al., 2023). We ignore the <10 s million-year span between an impact and life’s origin because it is small compared to our ∼500 million-year uncertainty for the timing of an OoL-favorable impact.

4. Discussion

Our estimated timing for biopoiesis is compatible with the earliest well-accepted geologic evidence of life in the form of stromatolites in the 3.5 Ga Pilbara block of western Australia (Walter et al., 1980; Buick et al., 1981; Van Kranendonk et al., 2018). Older geologic evidence of life exists, such as a >3.7 Ga black shale metamorphosed to graphite with negative ${\delta }^{13}$ C characteristic of biology (Rosing, 1999; Ohtomo et al., 2014). Also, Bell et al. (2015) discovered graphite inclusions in a 4.1 Ga zircon grain with ¹³C depletions compatible with biological activity, but this is a tentative biosignature because there are abiotic mechanisms for making isotopically light carbon (Javaux, 2019). If the Bell et al. (2015) finding is indeed a sign of life, then it would be compatible with our estimated timing in Figure 4 because we find that there is only a ∼10% chance of life starting after 4.1 Ga. In a similar vein, our results are broadly consistent with the recent ∼4.2 Ga phylogenetic estimate for the existence of the last universal common ancestor (Moody et al., 2024). In general, if life arose from post-impact-reducing atmospheres, the OoL was Hadean, not Archean.

Hadean zircons provide some constraint on Earth’s impact history, which we do not explicitly account for in our calculations. The oldest zircons are from Jack Hills, Australia, with U-Pb dates as old as ∼4.38 Ga (Valley et al., 2014). Benner et al. (2020) argued that a Moneta-scale impact ( $\sim 2\times {10}^{22}$ kg) could not have occurred after this date, because such a collision would heat the crust enough to potentially reset all U-Pb chronometers. Benner et al. (2020) made this claim based on the work of Abramov et al. (2013), who used models of impact ejecta coupled to simulations of radiogenic Pb-loss in zircons to investigate chronometers resetting during the hypothesized “late-heavy bombardment.” However, Abramov et al. (2013) considered impacts much smaller than Moneta (e.g., $<{10}^{21}$ kg) and did not clearly establish a minimum impactor mass that would reset all zircon chronometers globally. Given this ambiguity, our simulations do not use zircon chronometers as a constraint on Earth’s impact history.

Even if an impact failed to reset a zircon chronometer, the shock wave produced by an asteroid collision may create micro- to nanostructural features in zircons that could be preserved over billions of years (Reimink et al., 2023). Therefore, the presence or absence of shocked Hadean zircons in the geologic record may provide some constraint on Earth’s impact history. For example, Reimink et al. (2023) estimated the probability of preserved zircons with shock features assuming Earth experienced a “late-heavy bombardment” at 3.9 Ga. In this scenario, they found that shocked zircons were likely preserved, yet they found none in a collection of 4.02 Ga zircons from the Acasta Gneiss Complex. Overall, the absence of preserved shocked zircons suggests that a “late-heavy bombardment” did not occur at 3.9 Ga. Our Monte-Carlo models of Earth’s impact history do not use zircon shock features as a constraint. A better model would incorporate this information, noting that resolving the issue of preservational bias of zircons is beyond the focus and scope of this paper.

It is tempting to extrapolate our results beyond the early Earth to exoplanets, but we must do so with caution because planets orbiting different stars likely have bombardment histories, unlike the Hadean. Planets in the habitable zone of a late M-type star during the stellar main-sequence phase were interior to the habitable zone for several hundred million years during the superluminous pre-main-sequence phase (Luger and Barnes, 2015). Lichtenberg and Clement (2022) used N-body simulations to show that, for planets hosted by late M-dwarfs, large asteroid impacts are not useful for prebiotic chemistry because big impacts most likely occur within 100 million years of planet formation during the stellar pre-main-sequence when the planet is outside of the habitable zone. The pre-main-sequence phase is not an issue for an impact-induced OoL on planets that orbit sun-like stars because the phase only lasts ∼3 Myrs, while large asteroid impacts occur for 10⁸ years (Lichtenberg and Clement, 2022). Therefore, our result that an OoL-favorable impact was likely on early Earth (93% chance in an optimistic case) might be best extrapolated to habitable zone exoplanets orbiting sun-like stars.

A caveat to our results is that ocean vaporization by an impact is likely a complicated function of impact mass, velocity, and incident angle, which we do not fully account for (Appendix B). Throughout Section 3, we have assumed that 10% of an impactor’s kinetic energy heats the planetary surface ( $f_{\text{E},\text{vap}}=0.1$ ), leading to a required $>5\times {10}^{28}$ J collision to vaporize an ocean. Appendix Figure A4 uses simulations of impacts from the work of Citron and Stewart (2022) to show that $f_{\text{E},\text{vap}}=0.1$ is a reasonable assumption, but values between 2.5% and 25% are also possible, and $f_{\text{E},\text{vap}}$ likely also depends on impact angle.

To determine the sensitivity of our results to $f_{\text{E},\text{vap}}$ , we recomputed the likelihood and timing of an impact favorable for the OoL in Appendix B using $f_{\text{E},\text{vap}}=0.025$ and $f_{\text{E},\text{vap}}=0.25$ . We found that the probability of an OoL-favorable impact is somewhat sensitive to $f_{\text{E},\text{vap}}$ , while the timing does not depend strongly on $f_{\text{E},\text{vap}}$ . Overall, this uncertainty does not change our qualitative conclusion that an OoL-favorable impact on the early Earth was not a fluke. Even in our worst-case scenario (i.e., Wogan et al. (2023), a pessimistic case with $f_{\text{E},\text{vap}}=0.25$ ), an OoL-favorable impact occurs in 16% of impact realizations (Appendix Fig. A5). In addition, uncertainty in $f_{\text{E},\text{vap}}$ does not change our general conclusion that the most likely timing for the OoL was ∼4.35 Ga, with 95% uncertainty spanning the Hadean eon (Appendix Fig. A6). However, a more complete understanding of $f_{\text{E},\text{vap}}$ as a function of impact angle and other similar parameters may change this result.

An additional related caveat is that our calculations assume that an impact that vaporizes the ocean also sterilizes the planet, but this may not be the case if thermophilic microbes survived in the deep (>1 km) subsurface (Sleep and Zahnle, 1998; Zahnle and Sleep, 2006). If ocean vaporization did not destroy the biosphere, then our results would prefer an OoL earlier than we predict in Figure 4, and an OoL-favorable impact would also be more probable than we have estimated (Fig. 3).

5. Conclusions

We used Monte-Carlo simulations of Earth’s impact history to determine the most likely timing for the OoL, assuming it required abundant prebiotic nitriles. We used the results of Wogan et al. (2023), who found that significant OoL precursor molecules, such as HCN, were produced in the Hadean atmosphere after a $>4\times {10}^{20}$ kg impact in an optimistic case, and after a $>5\times {10}^{21}$ kg impact in a pessimistic scenario. We consider both possibilities and assume such impacts are favorable for life’s origin so long as they are not followed by an ocean-vaporizing collision that sterilizes the planet.

For either the optimistic or pessimistic cases, we found that the most likely timing for an OoL impactor is ∼4.35 Ga with a 95% uncertainty spanning the entire Hadean eon (approximately 4.45 to 3.9 Ga). The large uncertainty is caused by the intrinsic stochastic nature of impacts. These results suggest that Benner et al. (2020) proposed timing of the OoL, at 4.36 ± 0.1 Ga, is too narrow. Furthermore, the mass of the OoL-favorable impactor is most likely (80% to 96% probability) smaller than the Moneta impactor (Fig. 4) proposed by Benner et al. (2020), because the size-frequency distribution of asteroids favors more frequent smaller impactors.

Our simulations of Earth’s impact history do not always result in a bombardment favorable for an OoL. There are some impact histories where a collision of sufficient mass to produce prebiotic molecules does not occur or, alternatively, does occur but primitive life is subsequently destroyed by an ocean-vaporizing impact that does not rekindle the biosphere. With our nominal assumptions, an OoL-favorable impact occurs 93% or 30% of the time when we assume a $>4\times {10}^{20}$ kg or $>5\times {10}^{21}$ kg impact without later ocean vaporization required to produce substantial prebiotic molecules, respectively. In our worst-case scenario, which assumes that impacts can more easily vaporize the ocean, a $>5\times {10}^{21}$ kg impact without subsequent planet sterilization occurs in 16% of simulated impact histories (Appendix Fig. A5).

If life began after an impact that generated a reducing atmosphere, then this work suggests that OoL on Earth was fairly likely (at least a ∼16% chance) and indeed strongly favored (93% chance) under optimistic assumptions. Given that rocky planets form from accretionary impacts, our work supports an optimistic outlook for life on exoplanets that orbit sun-like stars and future searches for biosignatures.

6. Software

The source code needed to install the necessary software and reproduce all main text calculations (i.e., Figs. 2–4) is archived on Zenodo: https://doi.org/10.5281/zenodo.11051955 (Wogan, 2024).

A size-frequency distribution sensitivity test

We assume that the size-frequency distribution of objects that struck the early Earth is similar to the main-belt asteroids. A problem with this approach is that the main-belt asteroids only contain objects up to ∼1000 km diameter, yet the early Earth is expected to have experienced impacts larger than this (Marchi et al., 2014). Therefore, we must extrapolate the size-frequency distribution above ∼1000 km to larger objects. Throughout the main text, we use the same extrapolation as Marchi et al. (2014) (the black dashed line in Fig. 1), who extend the size-frequency distribution with a log-log slope of $d(\ln \hspace{0.17em}{S}_0)/d(\ln \hspace{0.17em}m)=-0.415$ .

To test the sensitivity of our results to this chosen extrapolation, we re-did our Monte-Carlo analysis using a log-log slope of $d(\ln \hspace{0.17em}{S}_0)/d(\ln \hspace{0.17em}m)=-1$ (the red dashed line in Fig. 1) for impactors bigger than ∼1000 km diameter. Figures A2 and A3 show how our results change when adopting this alternative extrapolation. Overall, our results are qualitatively unchanged for the different extrapolation slope ( $d(\ln \hspace{0.17em}{S}_0)/d(\ln \hspace{0.17em}m)$ ).

B impact energy required for ocean vaporization

Recently, Citron and Stewart (2022) performed smoothed-particle hydrodynamics impact simulations over a range of impact angles, masses, and velocities to estimate the impact properties that can vaporize an ocean. Figure A4a shows their simulation results for the change in the atmosphere’s and ocean’s internal energy ( $\Delta {\text{IE}}_{\text{atmo}}$ ) as a function of impact kinetic energy ( ${\text{E}}_{\text{imp}}$ ), along with log-log extrapolations for each impact angle. Figure A4 gives the same information, but the y-axis is, instead, the fraction of the impactor’s energy that heats the atmosphere and ocean over the course of their simulations. For the most probable incident impact angle of ${45}^{°}$ about ∼10% of the impactors kinetic energy heats the atmosphere ( $f_{\text{E},\text{vap}}=0.1$ ), so we nominally adopt this value in the main text to determine which collisions vaporize the ocean.

Energy delivery to the atmosphere/ocean appears to depend on impact angle, mass, and velocity (Fig. A4). In addition, all the Citron and Stewart (2022) simulations are far more massive than the minimum threshold for ocean vaporization, so we must rely on extrapolations. Therefore, our assumption of a constant $f_{\text{E},\text{vap}}=0.1$ is an oversimplification. To remedy this shortcoming, we considered building a parameterization for $f_{\text{E},\text{vap}}$ that depends on multiple impact properties (e.g., impact angle) by using the simulation results of Citron and Stewart (2022). This was, however, unsuccessful because their calculations do not consider a wide enough parameter space.

We evaluate the sensitivity of our results to an assumed constant $f_{\text{E},\text{vap}}=0.1$ , and we recomputed the likelihood and timing of an OoL-favorable impact using $f_{\text{E},\text{vap}}=0.025$ and $f_{\text{E},\text{vap}}=0.25$ (Fig. A5 and A6). We chose these values because they are reasonable lower and upper bounds based on the Citron and Stewart (2022) simulations (Fig. A4b). The probability of an impact that produces substantial prebiotic molecules without later ocean vaporization decreases with increasing $f_{\text{E},\text{vap}}$ (Fig. A5). For example, for $f_{\text{E},\text{vap}}=0.1$ , the probability of a $>4\times {10}^{20}$ kg impact (Wogan et al. (2023) optimistic case) without subsequent ocean vaporization is 93%. For $f_{\text{E},\text{vap}}=0.025$ and $f_{\text{E},\text{vap}}=0.25$ the probabilities are instead 99.9% and 55%, respectively.

Figure A6 shows the timing of the last impact to make conditions favorable for biopoiesis that does not experience subsequent ocean vaporization for $f_{\text{E},\text{vap}}$ values of 0.025, 0.1 (nominal value), and 0.25. As in the main text, we consider the Wogan et al. (2023) optimistic (Fig. A6a) and pessimistic (Fig. A6b) minimum impactor masses to produce significant prebiotic molecules. In Figure A6a, the timing of the last OoL-favorable impact is relatively insensitive to $f_{\text{E},\text{vap}}$ . In the Wogan et al. (2023) pessimistic case (Fig. A6b), an OoL is preferred later in the Hadean for larger values of $f_{\text{E},\text{vap}}$ .

Overall, uncertainty in the impact properties required to vaporize the ocean has a small effect on our qualitative conclusions. Regardless of $f_{\text{E},\text{vap}}$ , the OoL on Earth from an impact does not appear to be a fluke (Fig. A5), and biopoiesis is preferred early in the Hadean with uncertainty spanning the entire eon (Fig. A6). However, it is conceivable that a more complete understanding $f_{\text{E},\text{vap}}$ as a function of impact angle and other impact parameters could change these results.