The technique started on Earth. A 2017 paper used tidal tomography, the practice of reading a planet’s insides from the way its outside deforms under tidal pull, to constrain the buoyancy of Earth’s deep mantle. In 2025 the same approach was turned on the Moon, where a thermal asymmetry in the lunar mantle was inferred from its monthly tidal response.
Mars was the obvious next target and the awkward one. The pull that matters there is the Sun’s, and the component of it this method can read cleanly is the seasonal one, raised by Mars’s 0.0934 orbital eccentricity and its 25.2 degree axial tilt and varying across the Martian year of 687 Earth days.
Alexander Berne of the Lunar and Planetary Laboratory at the University of Arizona and colleagues now report in Nature that Mars flexes asymmetrically under that seasonal pull. Most of the stiffness contrast that flexing implies can be explained by mantle roughly 200 to 400 degrees Celsius warmer beneath the southern highlands than beneath the northern lowlands. The paper was received in December 2025 and accepted in July 2026.
Sixteen years of radio wobble
Why Mars has two faces is one of the older open questions in planetary science, and the paper describes its origin as widely debated. The northern hemisphere is low, smooth plain; the southern hemisphere stands higher, rougher and more densely cratered. The paper puts something like 25 kilometres of average difference in crustal thickness behind that, or alternatively a difference of roughly 200 kilograms per cubic metre in crustal density, and treats both as possibilities rather than settled numbers.
The interior was never imaged here. It was reconstructed from spacecraft motion. Berne and colleagues took X-band Doppler tracking collected by NASA’s Deep Space Network from three orbiters, Mars Global Surveyor, Mars Odyssey and Mars Reconnaissance Orbiter, covering about 16 years, and used the tiny changes in those spacecraft’s velocities to work out how Mars’s gravity field shifts through the Martian year.
The logic turns on a piece of symmetry. A perfectly spherically symmetric planet, squeezed by a tide, deforms only at the shape of the squeeze, which for solar tides is almost entirely what geophysicists call degree two. Such a planet would produce essentially no degree-three signal. Mars produces a large one: one coefficient departs from the spherically symmetric prediction by around 300 per cent, at better than 99.99 per cent confidence.
That figure is a residual. It survives only after the mass shuffled around by the Martian atmosphere has been modelled and subtracted, and the authors are candid that their error bars do not directly cover extreme atmospheric conditions. When they rerun the correction for a cold, dusty, low-solar-flux Mars, the north-south pattern still broadly aligns with the dichotomy, but of the two components they report for that case, one holds at three sigma and the other sits at two. They also note that solar tidal forcing is theoretically stronger at daily frequencies than at the seasonal one, and that they do not attempt to recover the daily signals at all, because the empirical accelerations they estimate to soak up unmodelled non-gravitational forces largely absorb them.
The stiffness map traces a line already drawn on the surface
Something inside Mars is not the same all the way round. To find out what, the team ran a Bayesian inversion over models in which the mantle’s effective shear modulus, its resistance to being sheared, varies from place to place.
The fit that works is hemispheric. Mantle beneath the northern lowlands is stiffer, centred near 45 degrees north and 138 degrees west over Vastitas Borealis; mantle beneath the southern highlands is softer, centred near 45 degrees south and 42 degrees east, close to the Hellas basin. The overall variation comes out at 81 per cent give or take 60 at three sigma, which is where the paper’s more cautious phrasing of more than 20 per cent comes from.
That centring is worth holding on to, because the announcement from Caltech, where Berne completed his doctorate before moving to Arizona, was headlined “Thermal Anomaly Discovered Below Mars’s South Pole”. Its own first line immediately widens the claim to the southern hemisphere, and the paper puts the softest mantle at 45 degrees south, a mid-latitude nearer Hellas than the pole. The pattern is hemispheric rather than polar: it spans the southern highlands, bounded by the dichotomy line rather than by latitude.
What makes the result more than a curiosity is where its boundary falls. The contour along which the stiffness difference passes through zero tracks the surface trace of the crustal dichotomy, including the way that boundary swings north in one hemisphere and south in the other. A structure inferred from radio wobbles lines up with a line mapped from topography. The correspondence weakens near the Tharsis volcanic rise, which the authors suggest may have overprinted the older boundary there.
The inversion also found no statistically significant lateral variation in the stiffness of the crust, and none at finer spatial scales than the hemispheric pattern within the mantle. Its amplitude is an average over the entire mantle depth range, 50 to 1,560 kilometres, because subdividing that range makes the answers trade off against one another and stop being significant. So the quantity here is a hemisphere-sized average rather than a bounded hot spot, though the authors add that this does not necessarily rule out more vertically localised structure, at the base of the lithosphere or near the core-mantle boundary.
From stiffness to degrees, and what the conversion cannot give
Rock softens as it warms, so a stiffness contrast is a candidate thermometer. The awkward part is that at the frequencies seismologists work with, producing a shear modulus difference this large would take a temperature contrast above 1,000 kelvin, which the authors call unrealistically large.
The length of the Martian year rescues the conversion. Extending laboratory measurements of olivine to an oscillation period of 687 days makes the mineral’s effective shear modulus roughly 15 to 20 times more sensitive to temperature than it is at seismic periods. Take the slice of parameter space that satisfies both the inferred stiffness contrast and Mars’s observed centre-of-mass offset, and most of the recovered variation corresponds to about 200 to 400 degrees Celsius. Independent thermal modelling by Plesa and colleagues had already predicted that a thick southern crust, acting as a blanket, could hold a hemispheric contrast above 200 degrees Celsius, though its figure is for a single depth rather than a whole-mantle average.
Composition cannot easily do the job instead. A purely compositional asymmetry heavy enough to explain the stiffness difference would displace Mars’s centre of mass from its centre of figure by roughly fifty times the observed offset. The models still allow up to about 5 per cent iron enrichment in the southern mantle alongside the heat, and that enrichment is itself a problem. The thermal scenarios that would warm the south, upwelling and insulation, involve melt extraction that should leave the southern mantle iron-poor, which is the opposite sign to the enrichment the inversion permits. A giant impact on its own, meanwhile, would not sustain a contrast of several hundred kelvin over a hemisphere for billions of years. So the authors sketch a hybrid origin as a way to resolve that mismatch, in which the Borealis impact that formed the northern lowland crust stripped iron from the northern mantle and the resulting thick southern crust then trapped heat underneath itself.
This is also where the second problem with the press treatment sits. Caltech’s release describes the southern interior as “partially molten.” The paper is more careful. Melt is not required by its results, and its cited marsquake data argue against a pervasively molten southern mantle: low-frequency events from the southern highlands give quality factors around 500, against 800 to 2,000 for events at Cerberus Fossae in the northern lowlands. Partially molten olivine would give values nearer 100 or lower, well below what the south returns. But the paper’s own conceptual figure does show melt forming below the lithosphere and stalling in the crust, and it names electromagnetic sounding as one future way to tell isolated magma pockets from a continuous molten layer at depth. Partial melting is a possibility the paper leaves open, not a finding it reports. It also bears on the headline number: the authors note that localised melting, if present, would lower the temperature contrast needed to explain the gravity data.
The temperature was not taken. It is an inference three steps removed from the data: spacecraft velocities give a time-varying gravity field, the gravity field gives a stiffness pattern, and a laboratory rheology for olivine converts stiffness into degrees. Each step carries assumptions, and the authors inflated their formal uncertainties fifteenfold before doing anything with them.
The volcanoes are in the wrong hemisphere
The paper does not need melt to make its case, but some thermal models predict that a warmer southern mantle would raise magma production there, and melt that forms at depth is expected to ascend and erupt over timescales of several million years. Yet Mars’s recent volcanism, within the last tens of millions of years, sits at Cerberus Fossae in the north.
The authors offer two ways out and do not choose between them. The thicker southern crust may block the ascent, leaving magma stalled in the middle crust as intrusions that would be hard to spot in surface images but might show up in very-high-resolution mapping of the static gravity field. Or eruption may need the ground to be pulling apart, which describes Cerberus Fossae and does not describe the southern highlands, where the tectonics are dominantly compressional.
A similar hedge sits around the planet’s magnetic record. Southern crust is more strongly magnetised than northern crust, and a warm southern mantle offers a story about how that happened while Mars still had a dynamo, between about 4.5 and 4.1 billion years ago. The mechanism has two stages and runs against intuition: upwelling would have heated the crust above the Curie temperature, erasing any magnetisation, and the rock would then have acquired the magnetisation it carries now on cooling back below that temperature while the dynamo was still running. The paper also offers a second route, a stronger regional dynamo driven by higher heat flow out of the core. It is a story the data permit rather than one they demonstrate.
Ganymede, Mercury, Io, Enceladus
The wider prize is not Mars. Tidal tomography relies largely on remote measurements, needing a spacecraft whose motion can be tracked precisely for long enough on top of a reference interior model, and the paper names Ganymede, Mercury, Io and Enceladus as bodies with the kind of large-scale lopsidedness it could read, calling the method a tool for characterising interiors without landed spacecraft. A mission built for the job, of the kind flown at Earth and the Moon, could recover the same structure in a fraction of the 16 years this result took, or resolve finer structure than a hemisphere.