Probing Mantle Geodynamics: Novel Constraints from Exoplanet Analogs

Mantle convection drives planetary evolution, but our understanding leans heavily on Earth. Seismic tomography offers snapshots of present-day structure; geoid anomalies and petrology give indirect clues about viscosity and temperature. What if we could test our models against planets with radically different sizes, compositions, and thermal histories? Exoplanet observations—especially radius–mass measurements and inferred bulk compositions of super-Earths and sub-Neptunes—now provide that chance. This article is for geoscientists who want to use exoplanet analogs to stress-test mantle dynamics models, uncover hidden assumptions, and build more robust frameworks that apply across a wider range of planetary conditions.

Why Exoplanet Analogs Challenge Conventional Mantle Dynamics

Terrestrial geodynamics assumes Earth's mantle behaves as a viscous fluid over geological time, with Rayleigh numbers around 10⁶–10⁸ and viscosity that depends strongly on temperature and pressure. Exoplanets with masses 2–10 Earth masses (super-Earths) have interior pressures reaching 1–10 terapascals—far beyond the stability field of bridgmanite. At these pressures, phase transitions to post-perovskite and potentially new silicate phases alter density, rheology, and heat capacity. The viscosity profile becomes non-monotonic, and the assumption of a single activation volume breaks down. Many super-Earths likely have stagnant lids (no plate tectonics) or episodic regimes, which changes the boundary condition for mantle convection. By confronting our models with exoplanet data—especially the radius–mass scatter that composition alone cannot explain—we can test whether our parameterizations of viscosity, thermal conductivity, and phase transitions are physically realistic.

One striking example: the observed radius gap (a relative dearth of planets with radii ~1.5–2 Earth radii) may reflect atmospheric loss, but it could also indicate a transition in mantle dynamics that affects outgassing and thermal evolution. Models that ignore pressure-dependent viscosity or post-perovskite transitions predict different cooling histories and thus different radius–age relationships. Exoplanet analogs thus act as a natural laboratory where we can vary gravity, core size, and composition in ways impossible on Earth.

The catch is that exoplanet observations are sparse and indirect. We have bulk density from radius and mass, but no seismic data. We must infer interior structure from evolutionary models that depend on uncertain initial conditions. This inverse problem is degenerate: different combinations of core mass fraction, mantle composition, and thermal state can produce the same radius. To break degeneracies, we need forward models that predict not just static structure but also observable signatures of dynamics—such as tidal heating, volcanic outgassing, and atmospheric composition. That is where the approaches described below come in.

What We Gain from the Exoplanet Perspective

First, we gain a wider parameter space: Rayleigh numbers can reach 10¹² or higher, and the viscosity contrast between top and bottom boundaries can exceed 10⁶. Second, we encounter new physics: radiative heat transfer becomes important at high temperatures, and the lattice thermal conductivity of post-perovskite is anisotropic. Third, we are forced to confront timescale issues: a 10-Earth-mass planet may have a convective overturn time of billions of years, meaning some planets may still be in a transient thermal state. These challenges push us to develop more general scaling laws that can be tested against Earth's own record.

Three Approaches for Incorporating Exoplanet Analogs

Researchers currently use three main strategies to connect exoplanet observations with mantle dynamics models. Each has distinct strengths and weaknesses, and the choice depends on the question being asked and the computational resources available.

1. Scaled Laboratory Experiments

Laboratory analog experiments use fluids with temperature-dependent viscosity (e.g., corn syrup or silicone oil) in a centrifuge to achieve high effective gravity. By adjusting the fluid properties and boundary conditions, one can simulate the high Rayleigh numbers and viscosity contrasts expected in super-Earths. These experiments provide direct visualization of flow patterns—plume spacing, boundary layer instabilities, and stagnant-lid thickness—that are difficult to resolve in numerical models. However, they are limited by the range of viscosity ratios achievable (typically ≤10⁵) and cannot easily incorporate compressibility or phase transitions. They are best suited for testing scaling laws for heat transport and mixing efficiency under extreme conditions.

2. Numerical Simulations with Generalized Equations of State

Numerical models remain the workhorse. The key innovation for exoplanet applications is the use of generalized equations of state (EOS) that cover a wide pressure–temperature range, such as the Vinet or Birch-Murnaghan EOS with parameters from first-principles calculations. These models can incorporate depth-dependent thermal expansivity, heat capacity, and thermal conductivity, as well as multiple phase transitions (e.g., olivine → wadsleyite → ringwoodite → bridgmanite → post-perovskite). The challenge is computational cost: 3D spherical-shell simulations at Earth-like Rayleigh numbers already require millions of CPU-hours; for super-Earths, the grid resolution must be finer to capture thin boundary layers. Many groups therefore use 2D axisymmetric or Cartesian geometry with parameterized convection (e.g., mixing-length theory) to explore parameter space quickly before running full 3D simulations.

3. Machine-Learning Emulators

Machine-learning emulators (e.g., Gaussian processes or neural networks) trained on a sparse set of high-fidelity simulations can predict convective heat flux, mixing efficiency, and thermal evolution across a continuous parameter space. This approach is particularly useful for Bayesian inversion of exoplanet radius–mass data: the emulator serves as a fast forward model that can be evaluated millions of times in a Markov chain Monte Carlo (MCMC) framework. The risk is that the emulator may extrapolate poorly outside the training region, especially if the underlying physics includes sharp phase transitions or bifurcations. Careful design of the training set—using Latin hypercube sampling or active learning—is essential.

Criteria for Choosing the Right Approach

Selecting among these methods depends on the specific research question, the available computational budget, and the need for direct observables. We recommend evaluating the following criteria:

• Parameter range: If you need to explore pressure >1 TPa or viscosity contrasts >10⁶, numerical models with generalized EOS are necessary; laboratory experiments cannot reach these extremes.
• Observable prediction: If your goal is to predict radius–mass evolution or tidal dissipation, an emulator coupled with an evolutionary code is most efficient. If you need flow patterns or mixing timescales, laboratory experiments or 3D simulations are better.
• Uncertainty quantification: Emulators naturally provide uncertainty estimates, but those uncertainties reflect only the training data. Numerical simulations can be run with multiple initial conditions to assess sensitivity, but this is costly.
• Reproducibility: Laboratory experiments are difficult to reproduce exactly; numerical models offer perfect reproducibility but may suffer from discretization errors. Emulators are deterministic once trained.
• Time to result: Laboratory experiments require setup and calibration (weeks to months). Numerical simulations can take days to weeks per run. Emulators, once trained, yield predictions in milliseconds.

For most research teams, we recommend a hybrid strategy: use numerical simulations with a generalized EOS to generate a training set, then build an emulator for parameter exploration and data inversion. Laboratory experiments can serve as validation for specific scaling laws (e.g., the relationship between Nusselt number and Rayleigh number) at moderate parameter ranges.

Trade-offs at a Glance: A Structured Comparison

The table below summarizes the key trade-offs across the three approaches, focusing on aspects most relevant to mantle dynamics research.

This comparison highlights that no single approach dominates. For example, if you are studying the onset of plate tectonics on a super-Earth, numerical simulations with a viscoplastic rheology are essential; laboratory experiments cannot reproduce the yield-stress behavior, and emulators may miss the bifurcation. Conversely, if you are trying to constrain the core mass fraction of a population of exoplanets, an emulator trained on a suite of thermal evolution models is the only practical way to perform Bayesian inference.

When to Avoid Each Approach

Laboratory experiments are a poor choice if you need to include radiative heat transfer or depth-dependent thermal conductivity. Numerical simulations may be overkill if you only need bulk heat flux scaling for a large parameter survey. Emulators should be avoided if the underlying physics includes discontinuous transitions (e.g., a sudden change in viscosity due to a phase change) unless the training set densely samples the transition region.

Implementation Path: From Model Design to Exoplanet Constraints

Once you have chosen an approach, the following steps outline a practical workflow for incorporating exoplanet analogs into mantle dynamics research.

Step 1: Define the Planetary Parameter Space

Start by selecting a range of planetary masses (e.g., 1–10 Earth masses), radii (from known exoplanet catalogs), and orbital distances (which affect tidal heating). Use a 1D interior structure model to estimate pressure, temperature, and density profiles for each candidate. This step provides the boundary conditions for the dynamics model. Pay attention to the core size: a larger core reduces mantle thickness and changes the aspect ratio of the convecting layer.

Step 2: Parameterize Viscosity and Rheology

Viscosity is the most sensitive parameter. Use a diffusion creep law with activation energy and volume from first-principles calculations or experimental data (e.g., for MgSiO₃ perovskite). For super-Earths, the activation volume may decrease at high pressure due to changes in defect formation energy. Implement a depth-dependent viscosity that includes the effect of the post-perovskite transition (which can reduce viscosity by a factor of 10–100). If plate tectonics is possible, add a yield-stress criterion (Byerlee's rule) to allow lithospheric failure.

Step 3: Choose the Convection Code and EOS

Open-source codes such as ASPECT, StagYY, or GAIA can be adapted for exoplanet conditions. Modify the material model to use a tabulated EOS (e.g., from the BurnMan library) that covers the pressure–temperature range of interest. Ensure the code can handle large density variations (up to a factor of 2–3 from surface to core–mantle boundary) and compressible convection. Test the code against benchmark cases (e.g., the Blankenbach et al. 1989 benchmark for isoviscous convection) before adding complexity.

Step 4: Run a Suite of Simulations

Design a factorial set of simulations varying mass, core size, reference viscosity, and internal heating rate. For each run, record the time-averaged Nusselt number, surface heat flux, and mixing efficiency (e.g., the mean age of mantle material). If using an emulator, this suite becomes the training dataset. Aim for at least 50–100 simulations to cover the parameter space, using Latin hypercube sampling for efficiency.

Step 5: Validate Against Exoplanet Data

Compare the predicted radius–mass relationship (including thermal expansion) with observed exoplanet radii from Kepler and TESS. Discrepancies may indicate missing physics (e.g., a different core composition, or a more insulating mantle). Also compare with atmospheric escape models: a mantle that cools faster will produce less outgassing, affecting the likelihood of a thick atmosphere. Use the emulator to perform Bayesian inversion to estimate the probability distribution of mantle viscosity, core size, and composition consistent with the observed population.

Step 6: Iterate and Refine

If the model fails to reproduce the observed radius gap or the scatter in radius at a given mass, revisit the viscosity law or include additional physics such as melting and magma ocean evolution. Exoplanet analogs are a moving target: as new data from JWST and PLATO become available, the constraints will tighten, requiring updates to the model suite.

Risks of Misapplying Exoplanet Analogs

The excitement of a new parameter space can lead to overinterpretation. Here are common pitfalls and how to avoid them.

Overfitting to Limited Data

Exoplanet radius–mass measurements have typical uncertainties of 5–10% in radius and 10–20% in mass. With only ~1000 well-characterized planets, it is tempting to tune model parameters to match every outlier. This overfitting produces models that fail to predict new observations. Mitigate by using cross-validation: hold out 20% of the data during training and test the model's predictive skill. Also, use Occam's razor: prefer models with fewer free parameters unless the data demand more complexity.

Misinterpreting the Stagnant-Lid Regime

Most super-Earths are expected to have stagnant lids because their high pressures suppress brittle failure. However, a stagnant lid does not mean no volcanism: plume melting can still occur, and the lid may be thin enough to allow some heat loss. Models that assume a mobile lid (plate tectonics) will overestimate cooling rates and underestimate mantle temperatures. Always check the surface boundary condition: if the yield stress is too high, the lid is stagnant; if too low, the model may produce unrealistic episodic subduction. Use a viscoplastic rheology with a yield stress that scales with pressure (e.g., 10–100 MPa at the surface, increasing with depth).

Neglect of Compositional Stratification

Exoplanets may have accreted heterogeneously, leading to a mantle that is not well-mixed. A dense basal layer (e.g., from impactor core material) can suppress convection and create a hidden reservoir that affects outgassing and thermal evolution. Models that assume a homogeneous mantle will predict different heat fluxes and mixing timescales. To test this, run simulations with an initial compositional gradient (e.g., a layer of iron-rich silicates at the base) and compare the thermal evolution with the homogeneous case.

Timescale Discrepancies

The convective overturn time for a super-Earth can be billions of years, meaning that many planets may not have reached a steady state. Using steady-state scaling laws (e.g., Nusselt–Rayleigh number relationships) derived from Earth's mantle may be invalid. Instead, run time-dependent simulations from initial conditions (e.g., a hot magma ocean) and track the evolution over the age of the system. The initial thermal state—whether the planet started hot or cold—has a lasting influence on the present-day radius and heat flow.

Frequently Asked Questions

How do we handle the uncertainty in exoplanet core size?

Core size is inferred from bulk density assuming a composition (typically iron and silicates). The uncertainty is large: a planet with a given mass and radius could have a core mass fraction ranging from 0.2 to 0.4. In mantle dynamics models, treat core size as a free parameter and run simulations across this range. The resulting spread in predicted heat flux and radius can be compared to observations to constrain core size statistically.

Can we use exoplanet observations to test the post-perovskite transition?

Indirectly. Post-perovskite has a higher thermal conductivity than bridgmanite, so its presence increases the efficiency of heat transport out of the core. This affects the thermal evolution and thus the radius–age relationship. If a population of planets shows a systematic deviation from models without post-perovskite, it could be evidence for the transition. However, the effect is subtle (radius change <1%) and may be masked by compositional variations.

What about tidal heating in close-in exoplanets?

Tidal heating can be a significant energy source, especially for planets on eccentric orbits. It depends on the mantle's tidal dissipation factor (Q), which is a function of viscosity and melt fraction. Incorporating tidal heating requires coupling an orbital evolution model with the mantle dynamics code. This is an active area of research, and the feedback between tidal heating and mantle convection can lead to thermal runaway or orbital circularization.

How do we validate a model without seismic data?

Validation relies on multiple indirect constraints: radius–mass relationship, atmospheric composition (from transmission spectroscopy), and orbital dynamics (e.g., tidal decay rates). Consistency across these independent datasets increases confidence. Additionally, models can be tested against Earth's own mantle by reducing the planet mass to 1 Earth mass and comparing with known seismic structure and heat flow.

Recommendations for Moving Forward

Exoplanet analogs are not a replacement for terrestrial geodynamics but a powerful complement. To make progress, we recommend the following specific actions:

• Adopt a hybrid modeling approach: Use numerical simulations with generalized EOS to build a training set, then develop an emulator for Bayesian inversion. This balances accuracy with computational feasibility.
• Include at least one non-Earth-like rheology: Test a stagnant-lid regime and a viscoplastic regime to see how the results change. Many exoplanets likely fall into one of these categories.
• Collaborate with exoplanet observers: Join the effort to characterize exoplanet atmospheres with JWST; the composition of outgassed species (e.g., CO₂, H₂O) can constrain mantle redox state and volatile content.
• Publish negative results: If a model fails to match observations, the failure is informative. It may indicate that the assumed viscosity law or phase diagram is incorrect, guiding future experimental work.
• Use open-source tools: Share your EOS tables, training data, and emulator code to accelerate the field. The ExoPlex and BurnMan libraries are good starting points.

The next decade will bring a flood of exoplanet data. By integrating these observations into mantle dynamics research, we can test the fundamental physics of convection under conditions far beyond Earth's, ultimately leading to a more robust understanding of our own planet's deep interior. Start by picking one of the three approaches above, adapt a convection code, and run a parameter sweep. The results will challenge your assumptions—and that is exactly the point.