Regolith Gardening: Micrometeorites Tilling an Airless Surface

The mechanism: a surface stirred by its own ejecta

An airless body has no shield. Dust that would flare and vanish 90 km above Earth's ground instead reaches the Moon at full orbital speed. The interplanetary mass flux at 1 AU peaks near particle masses of ~10^-5 g — grains 100-200 µm across — and delivers of order 10^3 tonnes per year across the Moon's 3.8x10^7 km^2. Spread evenly that is a derisory dusting. What matters is what each grain does on arrival.

• Comminution. A shock far above the target's strength shatters rock into fragments from millimetres down to sub-micron dust.
• Excavation. A crater tens to thousands of times the projectile's volume opens, and its contents fly out ballistically. On the Moon nearly all of it falls back: escape velocity is 2.38 km/s, well above typical ejection speeds.
• Overturn and burial. Ejecta lands inverted on neighbouring ground, burying what was exposed there, while the crater floor exhumes what was buried.
• Melting and welding. Part of the target melts and splashes over adjacent grains, gluing them into agglutinates — irregular, vesicular, glass-bonded aggregates found nowhere in terrestrial soil.

Repeat for 3.5 billion years, at every scale from micron pits to kilometre craters, and bedrock vanishes beneath a fragmental blanket in permanent slow motion. The blanket is the regolith; the turning-over is gardening. Both the name and the first quantitative models are Apollo-era: Gault, Hörz, Brownlee and Hartung's 1974 Mixing of the Lunar Regolith and James Arnold's 1975 Monte Carlo gardening model remain the reference framework.

The physics of a 20 km/s pinprick

Hypervelocity impact is an energy-density problem, not a momentum problem. Specific kinetic energy is just v^2/2: at 20 km/s that is 2x10^8 J/kg, about forty times the energy density of TNT (~4.6x10^6 J/kg). Heating silicate rock to its ~1500 K liquidus and melting it costs ~2x10^6 J/kg; vaporising it costs of order 1.4x10^7 J/kg. Impact at 20 km/s beats the melting budget ~100-fold and the vaporisation budget tenfold — which is why melt appears above roughly 5 km/s, vapour above roughly 10 km/s, and every lunar impact makes both.

Take one typical grain: a 100 µm sphere at 2.5 g/cm^3 masses ~1.3x10^-6 g and at 20 km/s carries ~0.3 J, the energy of dropping a 30 g weight one metre. Delivered into a spot narrower than a hair, it cuts a glass-lined "zap pit" about twice the projectile diameter, a spall zone several times wider, and throws out roughly 10^3 times its own mass. That amplification is the engine of gardening: the incoming flux is trivial, but the mass moved exceeds it by three orders of magnitude.

Two consequences follow. Regolith production is self-limiting — once the blanket is deeper than common impactors reach, craters merely reprocess it instead of making more, so thickness grows roughly as the square root of time, about 1 mm per million years averaged over mare history. And grain size reaches steady state: comminution grinds finer while agglutination welds fines back together, balancing at a median of ~45-100 µm in mature Apollo soils, with 10-20% of the mass below 20 µm.

Why the turnover clock depends so steeply on depth

Gardening rate is set by the crater size-frequency distribution. Write cumulative crater production as N(>D) ∝ D^-b, with b ≈ 3 over the centimetre-to-metre range on the Moon. A crater of diameter D disturbs an area ∝ D^2 and reaches a depth of roughly D/5, so overturning material at depth h requires craters with D ≳ 5h, and the rate at which those blanket a unit area goes as D^2 x D^-b = D^(2-b). The turnover timescale therefore scales as τ(h) ∝ h^(b-2) — roughly linear in depth for b ≈ 3, and steeper still where the small-crater branch steepens, because small craters outnumber large ones so overwhelmingly.

With real lunar numbers, the classic Gault-Arnold result is that the uppermost millimetre is reworked in ~10^4-10^6 years, a centimetre in ~10^6-10^7 years, and a metre only in ~10^8-10^9 years. A grain thus spends nearly all its life buried and inert, punctuated by brief violent visits to the surface. The samples agree: bulk mature mare soils have cosmic-ray exposure ages of only a few hundred million years although the surfaces are 3-3.8 Gyr old, and individual grains carry wildly different solar-flare track densities, some above 10^10 cm^-2 and others nearly track-free.

The modern revision came from Lunar Reconnaissance Orbiter. Emerson Speyerer and colleagues (Nature, 2016) compared thousands of LROC "temporal pairs" — before-and-after images of the same ground — and found 222 new craters ≥10 m in about seven years, plus more than 47,000 reflectance "splotches" from low-velocity secondary ejecta and impact jetting. Those secondaries churn the top layer far faster than primary-impact models allowed: the upper ~2 cm turns over in roughly 81,000 years, over 100 times faster than pre-LRO estimates. That number propagates straight into ice stability in polar cold traps and biosignature survival on icy moons.

Space weathering: agglutinates, nanophase iron, and why Tycho is bright

Gardening is the mechanical half; space weathering is the optical and chemical half. Impact vaporisation releases silicate vapour, and solar-wind hydrogen implanted in the target reduces its ferrous iron. Metallic iron condenses onto neighbouring grains as nanophase iron (npFe0) — blebs typically 3-33 nm across in amorphous vapour-deposited rims 60-100 nm thick. Lindsay Keller and David McKay imaged these rims directly by transmission electron microscopy of Apollo soil grains in 1993 and 1997, settling a decades-old argument about what darkens lunar soil.

The optical effect is size-dependent, a subtlety often missed. Iron particles far smaller than the wavelength (<~10 nm) are absorption-dominated in the Rayleigh limit, where extinction climbs steeply toward short wavelengths, so they suppress blue reflectance more than red — both darkening and reddening the spectrum. Larger "microphase" iron (≳50 nm), abundant inside agglutinitic glass, darkens broadly without much reddening — the distinction drawn by Sarah Noble, Carle Pieters and Keller in 2007. With agglutinates reaching 25-60% by volume, a mature surface shows lower albedo, a steeper red continuum slope, and suppressed 1 µm and 2 µm pyroxene bands. Bruce Hapke's 2001 radiative-transfer model, spanning "Mercury to the asteroid belt," is the standard treatment.

Maturity is measured, not guessed. Richard Morris's Is/FeO index — ferromagnetic resonance intensity of npFe0 normalised to total iron — sorts soils into immature (0-29), submature (30-59) and mature (60-100). From orbit the equivalent is the optical maturity parameter OMAT, derived by Paul Lucey and colleagues in 2000 from Clementine 750 nm reflectance and the 950/750 nm ratio. Together they explain the Moon's most conspicuous feature: rayed craters. Tycho (86 km, dated to 108 ± 4 Myr from exposure ages of Apollo 17 Station 6 boulders) and Copernicus (~800 Myr, from Apollo 12 KREEP glass) are bright because their ejecta is optically immature, not intrinsically white. Rays fade on a roughly billion-year timescale as gardening rebuilds npFe0 rims — hence thousands of ancient craters but only a handful of rayed ones.

Reading the archive: cores, tracks and returned samples

The regolith is a stratigraphic recorder, and Apollo drilled into it. Deep drill cores from Apollo 15, 16 and 17 reached ~2.2-3.0 m and revealed dozens of depositional units, each an ejecta blanket from some nearby event, with maturity oscillating up and down the column — direct proof that deposition is episodic, not steady. Thickness was measured seismically: the Apollo 16 Active Seismic Experiment found ~12 m at Descartes in the highlands against ~4-5 m at mare sites, matching what William Quaide and Verne Oberbeck inferred in 1968 from the concentric and central-mound shapes of small craters that just punch through to bedrock. Chang'e-4's Yutu-2 Lunar Penetrating Radar reproduced the result on the far side in 2020, imaging ~12 m of fines over stacked ejecta in Von Kármán crater.

Several independent clocks cross-check the rate. Etched solar-flare particle tracks record residence in the top few millimetres. Galactic cosmic rays make spallation nuclides — 21Ne, 38Ar, 3He, 26Al, 53Mn — down to ~1 m. Secondary neutrons peaking near 150-200 g/cm^2 are captured by 157Gd and 149Sm, shifting isotope ratios in proportion to time spent at metre depth. And solar wind, at ~3x10^8 ions cm^-2 s^-1 and ~1 keV/amu, saturates a grain's outer ~50 nm within ~10^4 yr of exposure, so bulk gas tracks total surface area ever exposed: helium-3, at solar-wind 3He/4He ≈ 4.5x10^-4, reaches ~1-30 ppb in mature ilmenite-rich mare soils (ilmenite retains helium best), for a whole-Moon inventory of order 10^6 tonnes.

The flux itself was calibrated in situ when Apollo 12 retrieved the Surveyor 3 television camera in November 1969 after 31 months on the surface, fixing the contemporary micrometeoroid rate from microcrater counts. Sample return then carried gardening beyond the Moon: JAXA's Hayabusa brought grains from Itokawa in 2010, and Takaaki Noguchi's team found lunar-style npFe0 rims on olivine — the first proof that S-type asteroids weather the same way. Hayabusa2 (Ryugu, 5.4 g, 2020), OSIRIS-REx (Bennu, 121.6 g, 2023) and Chang'e-5 and -6 (1,731 g in 2020; 1,935.3 g in 2024) have since broadened the inventory.

Where the simple picture fails

Gardening is not space weathering. The two are routinely conflated. Gardening is mechanical — stirring, burying, comminuting. Weathering is the chemical and optical alteration that gardening enables. A surface can be gardened normally yet weathered slowly if something blocks the solar wind, which is the leading explanation for lunar swirls like Reiner Gamma, bright sinuous markings sitting over crustal magnetic anomalies.

Not every body reddens. Dawn's 2011-2012 survey found essentially no lunar-style reddening on Vesta, even though its howardite-eucrite-diogenite crust is rich in FeO-bearing pyroxene — main-belt impacts average only ~5 km/s against the Moon's ~20, so far less melt and vapour is produced, and brisk regolith mixing and mass wasting keep refreshing the surface. Carle Pieters' team argued in 2012 that its brightness variations come instead from mixing with dark exogenic carbonaceous material. Carbonaceous asteroids can weather the opposite way — bluing and brightening — which badly complicates matching meteorites to parent bodies by spectrum alone.

Small bodies cannot hold their soil. Bennu's escape velocity is ~20 cm/s, so nearly all ejecta is lost and gardening should strip rather than build. That Bennu and Ryugu have regolith at all points to other comminution routes, notably thermal fatigue from ~200 K diurnal swings, which Marco Delbò and colleagues showed in 2014 can fracture rock on near-Earth asteroids faster than micrometeoroids do. OSIRIS-REx also found Bennu spontaneously ejecting particles, and on sampling in October 2020 discovered regolith so weakly cohesive the spacecraft sank in.

Levitating dust is oversold. Surveyor 7's 1968 "horizon glow" inspired decades of models in which photoelectric charging (the sunlit surface floats a few volts positive under solar UV) lofts a dense layer of micron dust. LADEE's Lunar Dust Experiment, flying 2013-2014, instead found a tenuous, permanently asymmetric cloud sustained by ordinary impact ejecta and peaking during meteoroid streams — no dense levitated population.

And the dust bites. Fractured in vacuum with no water or wind to round them, grains stay angular with fresh, chemically reactive surfaces. Apollo 17's Gene Cernan called dust "the most aggravating, restricting facet of lunar surface exploration"; it abraded suits, defeated seals, degraded radiators, and forced the crew to repair a torn rover fender with maps and tape. Harrison Schmitt's "lunar hay fever" prompted decades of toxicology work, and NASA now applies a permissible exposure limit of 0.3 mg/m^3 for a six-month mission.

Open questions

How fast is the top layer really stirred? The LRO secondary-driven figure of ~81,000 years for 2 cm is an order-of-magnitude jump over older models, and extrapolating it deeper is unsettled — yet it governs whether water ice survives in permanently shadowed polar craters and how deep Europa Clipper must look beneath a gardened, radiolytically cooked ice regolith.

How does gardening work on bodies too small to keep their ejecta — are Bennu-class regoliths built by impacts, thermal fatigue, or YORP-driven landslides? Can helium-3 mining ever close, when one tonne at 10 ppb means processing 10^8 tonnes of soil, a pit three metres deep over ~20 km^2, heated above 700 °C? And what sets the weathering rate on Mercury, where impact speeds average roughly twice lunar values (~40 km/s against ~20) and the solar wind is five to ten times denser? MESSENGER mapped the consequences; BepiColombo is built to measure the process.

Gardening is steeply depth-dependent: each decade deeper takes far longer to turn over, and each layer archives a different radiation clock.

Depth below surface	Time to overturn once	Dominant radiation record	How it is read
Top ~1 mm	~10^4-10^6 yr	Solar-flare heavy-ion tracks; solar-wind H and He implanted in the outer ~50 nm	Etched fossil tracks (>10^8 cm^-2); noble-gas release on stepwise heating
~2 cm	~8x10^4 yr (LRO secondaries)	Solar energetic particles; near-total solar-wind saturation	Before/after LROC image pairs; counting reflectance splotches
~10 cm	~10^7 yr	Solar cosmic rays fading out; galactic cosmic-ray spallation beginning	Cosmogenic 21Ne, 38Ar, 26Al depth profiles in cores
~1 m	~10^8-10^9 yr	Galactic cosmic rays and the secondary-neutron peak	157Gd and 149Sm neutron-capture shifts; 3-m Apollo drill cores
~5 m (base of mare regolith)	Longer than the age of the surface	Essentially shielded; unweathered bedrock	Active seismic profiling, Quaide-Oberbeck crater morphology, Yutu-2 radar