Mercury’s battered surface is more than a record of impacts and volcanism; it is a slow-motion ledger of a planet losing heat. The landscape—punctuated by long, curving cliffs and a proliferation of ridges—tells a process-oriented story: as Mercury’s iron-rich interior cools and changes phase, the rigid outer shell must accommodate a smaller volume. This article analyses that process step by step, traces how scientists infer a loss of planetary radius from surface structures, explores why estimates vary between roughly one and seven kilometres, and considers what upcoming observations will likely clarify.

The thermal history that sets the stage

Planetary cooling begins at formation. Mercury, like other rocky worlds, retained heat from the energy of accretion, from internal differentiation as dense iron sank to form a core, and from long-lived radioactive decay. But Mercury’s configuration is unusual: its metallic core occupies an enormous fraction of the planet’s radius—about 85 per cent—leaving a comparatively thin silicate shell of mantle and crust, roughly 400 kilometres thick. This geometry amplifies how core and mantle heat loss translate into changes in bulk volume.

Sources and sinks of heat

At least three contributions matter. First, the residual heat from formation dissipates over time. Second, as the core cools it can nucleate a solid inner core, and the phase change from liquid iron to solid reduces the occupied volume. Third, radioactive elements in the mantle and crust provide internal heating that declines as isotopes decay. Together these processes decrease Mercury’s internal energy and, with it, the physical volume of its interior.

Timescales and rates

Cooling is not a uniform process. Thermal models and geological evidence suggest a major episode of contraction early in Mercury’s history—billions of years ago—followed by a long tail of slower shrinkage. Geologically ‘recent’ shifts may still have occurred, but in planetary geology ‘recent’ spans tens to hundreds of millions of years. The key point is that most of the physical shortening happened in deep time, so no contemporary instrument directly observed Mercury’s radius change; scientists read it from the planet’s wrinkles.

How a cooling interior produces cliff-like scarps

On bodies with mobile plates like Earth, compression can be accommodated at plate boundaries. Mercury lacks that plate tectonics system. Its lithosphere behaves broadly as one rigid shell. When the interior volume declines, the shell must shorten laterally. That shortening is achieved by thrust faults: one block of crust is shoved up and over an adjacent block. Where such thrusts break the surface, they form steep, often arcuate scarps called lobate scarps.

Geometry and scale of lobate scarps

These scarps can be immense. Orbital imagery and topography reveal examples hundreds to a thousand kilometres long and rising a kilometre or more—some local relief measurements even exceed three kilometres. They cut across older craters and volcanic plains, indicating that contraction occurred after much of the surface was emplaced. The curvature of these scarps is a direct manifestation of global shortening distributed across a brittle shell.

Why scarps are called wrinkles

The analogy of a sheet wrinkling as it shrinks is helpful: length must be accommodated within a smaller circumference, so the sheet forms folds and thrusts. On Mercury the scales are planetary; the wrinkles are physical cliffs that preserve how much shortening the lithosphere has absorbed.

From images to numbers: inferring global contraction

No instrument measured Mercury’s radius at two separate ancient epochs and directly subtracted one from the other. Instead, scientists infer global contraction by measuring shortening recorded in fault geometries, counting and mapping contractional landforms across the globe, estimating their cumulative displacement, and scaling those values into an equivalent change in planetary radius. Each step in this inference chain involves assumptions and choices that affect the final answer.

Early measurements and the MESSENGER revolution

Mariner 10’s flybys in the 1970s imaged less than half the planet, yielding preliminary estimates of radius reduction around one to two kilometres. The MESSENGER mission (orbiter operations 2011–2015) transformed the data set: near-global images, stereo mapping, and topographic information made it possible to inventory thrust faults and ridges across almost the entire planet. That fuller catalogue enabled more ambitious scaling and higher estimates of overall shortening.

Counting faults, measuring displacement

To convert observed scarps into a global radius change, researchers measure ridge length, estimate displacement across individual thrusts, and sum these measurements. Some features, like high-relief lobate scarps that break ancient terrain, are clear markers of global-scale thrusting. Others, such as wrinkle ridges in volcanic plains, can be ambiguous: they might represent broad contractional folding or instead be the product of local subsidence, loading, or shallow flexure. Whether to include these features in the tally is a major methodological hinge.

Why scientists disagree: methodological forks in the road

Two high-profile analyses illustrate the impact of different cataloguing philosophies. A 2014 study by Paul Byrne and colleagues used an expanded inventory of contractional structures, including long belts of ridges and smaller features. Scaled globally, their measurements were consistent with as much as seven kilometres of radius reduction. That figure entered popular accounts as the headline shrinkage number.

A narrower catalogue and a smaller planet

In 2021 Thomas Watters presented a contrasting approach. He argued that many small positive-relief structures could not be demonstrably tied to deep, global thrusting and that a conservative inventory—assigning one principal fault to clearly contractional landforms and excluding ambiguous or local features—better constrained global shortening. His analysis produced a much smaller estimate: about one to two kilometres of radius reduction.

The role of interpretation

The core of the disagreement is interpretive: do wrinkle ridges in volcanic plains record a planetary-scale shortening component, or do they mostly record local mechanical responses? Each interpretation can be defended, and each choice affects the summed displacement and therefore the global contraction estimate. Because these structures cannot be excavated, and the fault geometries at depth are not directly observed, scientists must rely on surface expression, context, and modeling assumptions.

Machine learning, a middle path, and temporal constraints

A 2026 study by Adrien Broquet and Jeffrey Andrews-Hanna used machine-learning-assisted analysis of topography to try to reduce subjective biases. Their method estimated ridge heights, filtered out short secondary features near longer primaries, and accounted for fault directionality. When wrinkle ridges were included the model yielded about 6.3 kilometres of global contraction; when they were excluded it fell to about 1.2 kilometres. The paper also inferred a temporal concentration of contraction between roughly 4.1 and 3.9 billion years ago, with much lower rates afterward.

What that result tells us about the debate

The 2026 analysis is important because it shows that methodological refinement can push estimates back toward the higher end, while also making explicit how different catalogues change the outcome. It exposes the hinge: the treatment of smaller ridges. If those features record distributed global shortening, Mercury lost several kilometres in radius. If they predominantly record local flexure and subsidence, the true contraction is closer to a kilometre or two.

Implications: what the amount and timing of shrinkage mean

The magnitude of radius loss constrains Mercury’s thermal history and interior evolution. Losing six to seven kilometres implies a larger total heat loss and a substantial reorganisation—perhaps earlier and more significant inner-core solidification—than a scenario where only one or two kilometres were lost. The timing matters for models of how and when Mercury’s dynamo generated a magnetic field and for reconciling observed present-day, weak magnetism with the planet’s small size.

Late-stage activity and the misconception of a present-day shrink

Although most contraction happened long ago, some studies find evidence that tectonic movement persisted later than the main early pulse. For example, small grabens preserved atop larger contractional structures suggest episodes of extension after contraction, interpreted as evidence for prolonged thermal and tectonic evolution. But these findings do not equate to live shrinking observable on human or even astronomical timescales; ‘recent’ in this context can still span hundreds of millions of years.

What BepiColombo will contribute and what to watch for

The joint ESA-JAXA BepiColombo mission, arriving at Mercury in late 2026 with full science operations beginning in 2027, promises a new level of constraint. Two orbiters carrying high-resolution stereo cameras, laser altimetry, gravity instruments, and magnetometers will improve topography, detect subtle displacements, and refine crustal structure models. Those data will allow researchers to measure ridge heights and displacements more accurately, map subtle fault scarps, and better discriminate between global thrusting and local deformation mechanisms.

Specific observations that will reduce ambiguity

Key advances to expect include: stereo-derived topography at higher resolution than MESSENGER’s, enabling clearer displacement estimates; altimetry profiles that can measure scarp relief perpendicular to their strike; higher-resolution gravity data to probe crustal and mantle structure; and magnetometry that can inform models of inner-core solidification timing. Together these data types will tighten constraints on the total shortening and on when most contraction occurred.

How to interpret future refinements

Even with better data, some interpretive choices will persist. Improved resolution reduces, but does not eliminate, ambiguities in assigning faults and measuring displacements. The most robust path forward is integrative: combine morphological mapping, high-resolution topography, gravity inversions, and mechanical modeling. Machine learning and automated feature detection can help by applying consistent criteria across data, but human geological judgment will remain necessary to interpret context and process.

Mercury’s long cliffs are a planetary palimpsest: each scarp, ridge and trough is a layer of evidence left by a cooling, contracting interior. The debate over whether the planet lost only a single kilometre or several thousand metres of radius is not a contest about whether Mercury shrank—the planet clearly did—but about how much of the surface record should be counted as evidence of global shortening versus local adjustments. As BepiColombo’s data arrive and analyses grow more refined, we should expect the story to move from plausible ranges to tighter constraints, and with them a clearer account of how a small world adjusted to the slow, inevitable loss of its internal heat.