UC Berkeley

Rotating planets are most stable when spinning around their maximum moment of inertia, and will tend to reorient themselves to achieve this configuration. Geological activity redistributes mass in the planet, making the moment of inertia a function of time. As the moment of inertia of the planet changes, the spin axis shifts with respect to a mantle reference frame in order to maintain rotational stability. This process is known as true polar wander (TPW). Of the processes that contribute to a planet's moment of inertia, convection in the mantle generates the largest and longest-period fluctuations, with corresponding shifts in the spin axis. True polar wander has been hypothesized to explain several physiographic features on planets and moons in our solar system. On Earth, TPW events have been invoked in some interpretations of paleomagnetic data. Large swings in the spin axis could have enormous ramifications for paleogeography, paleoclimate, and the history of life.

Although the existence of TPW is well-verified, it is not known whether its rate and magnitude have been large enough for it to be an important process in Earth history. If true polar wander has been sluggish compared to plate tectonic speeds, then it would be difficult to detect and its consequences would be minor. Herein I investigate rates of true polar wander on convecting planets using scaling, numerics, and inverse problems.

I perform a scaling analysis of TPW on a convecting planet, identifying a minimal set of nondimensional parameters which describe the problem. The primary nondimensional numbers that control the rate of TPW are the ratio of centrifugal to gravitational forces m and the Rayleigh number Ra. The parameter m sets the size of a planet's rotational bulge, which determines the amount of work that needs to be done to move the spin axis. The Rayleigh number controls the size, distribution, and rate of change of moment of inertia anomalies, all of which affect the rate of TPW. I find that the characteristic size of moment of inertia anomalies decreases with higher Ra, but that the characteristic response time for TPW also decreases. These two effects approximately cancel. However, the orientation of the principal axes of the moment of inertia becomes less stable to perturbations at high Ra, thereby increasing the rate of TPW. Overall, I find that a more vigorously convecting planet (one with a higher Ra) is more likely to experience large TPW events. If early Earth had more vigorous convection, it may have experienced more TPW than present-day Earth.

Flow induced by density anomalies in the mantle deflects free surfaces at the surface and the CMB, and the mass anomalies due to these deflections contribute to the moment of inertia. A full accounting of the moment of inertia anomalies must include these surface effects. Numerical models of mantle convection with a free surface have suffered from numerical sloshing instabilities. I analyze the sloshing instability by constructing a generalized eigenvalue problem for the relaxation time spectrum. The minimum relaxation time of the spectrum sets the maximum stable timestep. This analysis gives the first quantitative explanation for

why existing techniques for stabilizing geodynamic simulations with a free surface work. I also use this perspective to construct an alternative stabilization scheme based on nonstandard finite differences. This scheme has a single parameter, given by an estimate of the minimum relaxation time, and allows for still larger timesteps.

Finally, I develop a new method for analyzing apparent polar wander (APW) paths described by sequences of paleomagnetic poles. Existing techniques, such as spline fits and running means, do not fully account for the uncertainties in the position and timing of paleomagnetic pole paths. Furthermore, they impose regularization on the solution, and the resulting uncertainties are difficult to interpret. Our technique is an extension of paleomagnetic Euler pole (PEP) analysis. I invert for finite Euler pole rotations that can reproduce APW paths within a Bayesian Markov chain Monte Carlo (MCMC) framework. This allows us to naturally include uncertainties in age and position, and provides error estimates on the resulting model parameters. Regularization can be accomplished via

physically motivated choices for the parameters' prior probability distributions.

I applied the Bayesian PEP technique to the Mesoproterozoic Laurentian APW track, which primarily comes from the Keweenawan Midcontinent Rift. I fit the track with one and two Euler rotations. Both inversions did a good job of reproducing the Keweenawan track, though the two Euler pole inversion has a closer fit. I find that the implied Laurentian plate speeds exceeds 22.9 cm/yr at the 95\% confidence level. These speeds are significantly faster than Cenozoic plate speeds, and could be explained by either faster plate speeds in the Proterozoic or a TPW event.