An Alberta model traces decadal day-length swings to the inner core's gravitational pull

The solid inner core turns independently inside the molten outer core, and both sit within a ball of mostly iron roughly the size of Mars ^[22]. Averaged over time, the bumps on the inner core's boundary line up with the dense patches of mantle that pull on them, offset slightly to the east ^[13]. East-west flows in the outer core exert electromagnetic stress on the inner core and twist it out of that alignment, and the mantle feels a fluctuating gravitational torque in return ^[14]. Dumberry said the inner core "wants to be aligned" with the mantle because its dense regions are attracted to the mantle's dense regions, and that the generally westward flows of the outer core drag it out of line before gravity pulls it slowly back eastward ^[20].

The two rival torques each had a numerical problem. Viscous stress from core flows is far too small to do the job ^[8]. Electromagnetic coupling can reach the required magnitude if a several-kilometre iron-enriched layer with a conductance near 10^8 siemens lies at the base of the mantle, which studies of Earth's nutations have suggested ^[9], but predicted electromagnetic torques generally match the observed decadal record poorly unless the core flows are designed to produce the torque needed ^[10]. Topographic predictions built from geostrophic pressure and seismically derived maps of core-mantle boundary topography come out about 100 times too large, and they swing with small changes to the topography or the flow ^[11].

Earth-like dynamo models produce the same competition between the three torques that Zhang and Dumberry reconstruct ^[4]. "Before we obtained the result, we didn't know they are competing with each other," Zhang said ^[18].

Their record is short relative to the signal. Fluctuations of a few milliseconds run on periods of 10 to 70 years, and the series shown covers 1964 to 2019 ^[1]. That is 55 years, so the slowest component in the band never completes one cycle inside the data ^[23]. Only about nine years of the window fall after the 2010 turnaround in inner core rotation ^[24].

The same reconstruction also yields an inner core soft enough to deform within a few years and an upper limit on how strongly the top of the fluid core is stratified ^[7]. The authors write that their torque histories are tied to the accuracy of the inner core rotation and core flow models ^[5]. The signal underneath all of it is a few milliseconds of Earth's spin, measured by observing the planet's position relative to distant cosmic objects ^[21].