Earth may be able to maintain liquid-water oceans and a climate similar to its preindustrial state even without life, according to a computer model that traces the evolution of the planet’s interior, oceans, atmosphere and climate over 4.5 billion years. The simulated planet reaches a surface temperature of 286.9 kelvins, close to the preindustrial Earth, but lacks the abundant atmospheric oxygen and methane associated with modern life. The results suggest that a planet can remain habitable without being inhabited, although important uncertainties remain about how an entirely lifeless world would regulate its ocean chemistry and carbon cycle.
Earth has maintained conditions suitable for liquid water over much of its history, despite changes in the Sun’s brightness, the planet’s interior and its atmosphere. Life has profoundly influenced the modern environment, but it remains uncertain whether biological activity is necessary to maintain long-term habitability.
To investigate that question, Samantha Gilbert-Janizek of the University of Washington and colleagues developed a computer model of Earth evolving without life. The team followed the planet from 50 million years after its formation through 4.5 billion years of evolution, simulating the interactions between its core, mantle, crust, oceans, atmosphere and changing Sun.
The central question was whether a lifeless Earth could eventually develop conditions resembling those of the preindustrial planet, before large-scale industrial activity altered the climate.
The researchers defined habitability in terms of the conditions needed to maintain liquid water on the planetary surface. They did not attempt to reproduce every aspect of Earth’s history or demonstrate that life could never have originated. Instead, they tested whether a modeled planet without biological activity could reproduce a broad set of observed properties of the preindustrial Earth.
Their model matched 19 key characteristics spanning the atmosphere, climate, ocean chemistry and deep interior. Seventeen fell within their reported measurement uncertainties. Two ocean chemistry measurements, dissolved carbon dioxide and pH, differed modestly from their target values.
The result suggests that life may not be essential for maintaining Earth-like surface temperatures over geological timescales. It does not establish that every lifeless rocky planet can remain habitable, or that the actual early Earth followed the simulated history.
The planet’s interior helps control its surface
A planet’s atmosphere cannot be understood in isolation from the processes occurring beneath its surface. The movement of material through the mantle, volcanic activity, the formation of oceanic crust and the recycling of material at tectonic plate boundaries all affect how carbon and water move between the interior, oceans and atmosphere.
Those exchanges influence the concentration of atmospheric carbon dioxide, a greenhouse gas that helps determine surface temperature.
The team built its model using VPLanet, incorporating existing descriptions of the carbon cycle, water cycling, ocean chemistry and climate. It also included the thermal evolution of the mantle and core, the generation of Earth’s magnetic field, and changes in the energy received from the Sun.
The model calculates how heat moves through the interior as radioactive elements decay and the core and mantle cool. It tracks how mantle convection drives plate motion and how that motion influences volcanic outgassing, the return of water and carbon to the interior, and the chemical weathering of surface rocks.
The researchers also incorporated wavelength-dependent radiative transfer to calculate the balance between incoming stellar energy absorbed by the planet and outgoing heat released to space. This approach accounts for how atmospheric gases absorb and scatter radiation at different wavelengths, rather than treating the atmosphere as a single, uniform layer.
That combination matters because the planet’s surface temperature depends on more than the amount of sunlight it receives. Atmospheric composition, water vapor, surface reflectivity and clouds all affect the energy balance.
The model therefore evolves the planet as an interconnected system rather than treating its climate, interior and atmosphere as separate problems.
Carbon dioxide initially builds up, then declines
The simulated Earth begins its evolution after the solidification of its early magma ocean, at an age of about 50 million years. The researchers initialized the surface with two terrestrial oceans’ worth of water and a total initial carbon dioxide inventory of approximately 1.1 × 10²¹ kilograms.
They assumed active plate tectonics throughout the modeled history. They also set the atmosphere to contain one bar of nitrogen and excluded biological sources of atmospheric oxygen and methane.
These choices allowed the researchers to investigate whether an abiotic carbon cycle could sustain habitable surface conditions without relying on biological greenhouse gases.
Early in the simulation, volcanic and other modeled interior processes release carbon dioxide into the atmosphere. Following an initial equilibration period of roughly 100 million years, the surface temperature is about 297 kelvins. The atmospheric carbon dioxide pressure rises to approximately one bar within the first 100,000 years after model initialization.
As the surface warms, weathering becomes more effective. Chemical reactions between atmospheric carbon dioxide, water and rocks transfer carbon into surface reservoirs and ultimately help remove it from the atmosphere.
In the model, these carbon sinks gradually become stronger than the sources supplying carbon dioxide. Atmospheric carbon dioxide consequently declines as the planet evolves.
This process is part of the carbonate-silicate cycle, a set of geological and chemical exchanges that can regulate atmospheric carbon dioxide over long periods. The modeled cycle operates without biological activity, although its behavior depends on the assumptions built into the simulation.
By 4.5 billion years, the modeled atmosphere has a carbon dioxide partial pressure of 28.2 pascals, close to the preindustrial target of 28.4 ± 0.4 pascals.
The team also calculated a total carbon dioxide outgassing flux of about 9 trillion moles per year at the end of the simulation, reasonably close to the present-day observational estimate used for comparison.
The results indicate that the modeled geological carbon cycle can reduce atmospheric carbon dioxide to approximately preindustrial levels without requiring life to drive the process.
A habitable climate under a gradually brightening Sun
The Sun’s brightness changes over its lifetime, creating another challenge for a planet attempting to maintain liquid water. The model accounts for the Sun’s increasing luminosity as Earth evolves.
At the beginning of the simulation, the incoming solar flux is about 943 watts per square meter, approximately 70% of its present-day value. By 4.5 billion years, it reaches about 1,370 watts per square meter.
The planet’s surface temperature changes in response to this evolving energy supply and to changes in atmospheric composition and reflectivity.
By the end of the simulation, the model produces a global average surface temperature of 286.9 kelvins, or about 13.8 degrees Celsius. This matches the preindustrial temperature target used by the researchers.
The modeled ground albedo, a measure of how much incoming sunlight the surface reflects, is 0.14. When clouds are included, the total albedo rises to 0.198.
The simulation also produces an atmospheric water-vapor partial pressure of 1,600 pascals, compared with a preindustrial target of 900 ± 700 pascals.
Water vapor increases as the surface warms because warmer air can support more atmospheric moisture. Since water vapor also absorbs outgoing heat, this creates a feedback between temperature and atmospheric composition.
The researchers’ cloud treatment is simplified, however. They adjusted cloud opacity to reproduce the preindustrial surface temperature, and the resulting model corresponds to a cloud fraction of only about 8%. They note that this is probably unrealistically low for the preindustrial Earth, reflecting limitations in how clouds are represented.
The agreement in surface temperature therefore demonstrates that the modeled system can reproduce the chosen climate target, not that every detail of a lifeless planet’s climate has been independently established.
The oceans change as carbon moves through the system
The model follows water and carbon as they move between the mantle, oceans and atmosphere. This provides another test of whether a lifeless Earth can approach the observed preindustrial state.
Over 4.5 billion years, approximately one of the two initial terrestrial oceans’ worth of water enters the mantle, leaving one ocean’s worth at the surface. The simulated final ocean mass is 1.4 × 10²¹ kilograms, matching the target value.
Ocean chemistry evolves alongside atmospheric carbon dioxide. When atmospheric carbon dioxide is high, more dissolves in seawater, increasing the amount of dissolved inorganic carbon. This includes dissolved carbon dioxide, bicarbonate and carbonate.
Early in the simulation, the ocean becomes much more acidic than the preindustrial ocean. After the initial equilibration period, the modeled pH is about 6.6 as atmospheric carbon dioxide rises. As carbon dioxide subsequently declines, the ocean becomes increasingly alkaline.
By the end of the simulation, the modeled total dissolved inorganic carbon concentration at the ocean surface is 1,919 micromoles per kilogram, compared with the target of 2,000 ± 200 micromoles per kilogram.
The modeled ocean pH reaches 8.28, slightly more alkaline than the preindustrial target of 8.18 ± 0.05. The modeled dissolved carbon dioxide concentration is 10.8 micromoles per kilogram, compared with a target of 8 ± 2 micromoles per kilogram.
Both ocean chemistry values fall outside their stated measurement uncertainties, although the authors regard the discrepancies as modest.
The researchers attribute these differences in part to their simplified treatment of marine calcite, a mineral involved in ocean carbon chemistry. Their model assumes calcite saturation and tunes the calcite concentration to reproduce the observed dissolved inorganic carbon. It does not independently track total alkalinity, which also helps determine ocean pH and the distribution of dissolved carbon species.
That limitation is important because the model’s ability to reproduce a habitable climate depends partly on how effectively carbon can be transferred into ocean and geological reservoirs. Whether an entirely lifeless ocean could accumulate and maintain the calcite concentrations needed for this process remains unresolved.
The cooling core changes the planet’s magnetic field
The simulation also follows the evolution of Earth’s deep interior, including the mantle’s heat loss, the core’s cooling and the formation of the solid inner core.
These processes provide additional tests of the model beyond surface temperature and atmospheric composition.
By the end of the simulation, the upper mantle reaches a modeled temperature of 1,561 kelvins, compared with the observational target of approximately 1,587 kelvins. The temperature at the core-mantle boundary reaches 3,889 kelvins, within the target range of 4,000 ± 200 kelvins.
The model also reproduces the upper-mantle heat flow of 38 terawatts and produces a core-mantle boundary heat flow of 15 terawatts, compared with a target of 11 ± 6 terawatts.
Other modeled properties include mantle viscosity, the fraction of mantle material that melts, the rate of melt production and the radius of the inner core. These results fall within the observational uncertainties used for the model’s validation.
One notable change occurs at about four billion years in the simulation, when the inner core begins to solidify. As iron crystallizes, lighter elements are released into the surrounding liquid outer core. This adds compositional buoyancy, changing the modeled energy available to drive motion in the core.
The simulation consequently produces a change in the core’s buoyancy flux, accompanied by changes in the magnetic moment and magnetopause radius. The final modeled values for the magnetic moment and magnetopause radius are close to their respective observational targets.
These results demonstrate that the coupled model can reproduce several present-day properties of Earth’s interior while also tracking their evolution over billions of years. They do not independently establish that the actual core followed precisely the same sequence of events.
A lifeless Earth would have a different spectrum
The researchers also investigated how the modeled planet might appear to a future telescope observing its reflected light.
They used the PICASO radiative-transfer code to generate synthetic spectra covering wavelengths from 0.2 to 2.0 micrometers. One spectrum represents a realistic, cloudy modern Earth. The other represents the modeled abiotic Earth after 4.5 billion years.
Both atmospheres contain water vapor, carbon dioxide and nitrogen. The key difference is that the abiotic atmosphere lacks oxygen, ozone and methane.
Those missing gases produce distinct differences in the spectra.
In the modeled modern Earth spectrum, ozone absorbs ultraviolet light near 0.25 micrometers and produces another absorption feature around 0.6 micrometers in visible light. Molecular oxygen produces a prominent feature near 0.76 micrometers and a weaker feature around 1.27 micrometers in the near-infrared.
Methane, present at about two parts per million in the modern Earth reference atmosphere, produces a comparatively weak absorption feature near 1.69 micrometers. That feature is absent from the abiotic spectrum.
Both spectra retain absorption features associated with water vapor and carbon dioxide.
The comparison demonstrates how a modeled planet with broadly Earth-like surface conditions could have an atmospheric spectrum different from modern Earth’s because it lacks biologically associated gases. It also illustrates why evidence of liquid-water habitability alone cannot establish that a planet is inhabited.
The study uses these spectra as a proof of concept for the planned Habitable Worlds Observatory, which is intended to characterize Earth-like planets around Sun-like stars. The simulated observations are not measurements of an actual exoplanet, and the study does not establish that any particular spectral feature will be detectable with the future telescope.
Important uncertainties remain
The researchers emphasize that their results depend on assumptions that simplify several aspects of planetary evolution.
The model assumes that plate tectonics operated throughout the simulated history. The actual timing and nature of the transition to modern plate tectonics remain uncertain, and the study does not attempt to reconstruct the tectonic regime of the earliest Earth.
The model also does not fully represent the atmospheric chemistry of a lifeless planet. Methane is excluded, even though the authors note that small amounts of abiotic methane could affect atmospheric chemistry and climate. Its possible accumulation would depend on processes not fully captured by the current model, including volcanic methane production, the chemical state of the mantle and atmospheric photochemistry.
The ocean chemistry assumptions introduce another uncertainty. Because calcite saturation is imposed rather than calculated from an independently evolving total alkalinity, the model cannot fully establish whether an abiotic ocean would maintain the chemistry needed to regulate atmospheric carbon dioxide over geological timescales.
Clouds are also treated in a simplified way, and the researchers do not validate the model against the uncertain conditions of the early Earth. Instead, they calibrate it to the preindustrial Earth, whose properties are better constrained across the different parts of the planet.
To reproduce those properties, the team optimized 16 initial conditions and model constants, using 200 starting points for its parameter optimization and selecting the best-fitting solution among runs that converged. The resulting agreement with present-day observations provides a test of the model, but it does not eliminate uncertainties in the assumptions or establish a unique history for Earth.
The authors propose that further development should incorporate more complete ocean chemistry, investigate abiotic methane and oxygen, and explore how different planetary compositions and host stars affect long-term habitability. They also identify the need to generate spectra for a broader range of evolving lifeless planets before the model can provide a more comprehensive baseline for interpreting possible signs of life.
The study was published in The Planetary Science Journal.
