Terminal descent, engine plumes and the Martian landing zone

The final twenty metres: when the engines begin to move Mars

For most of descent, the ground is a geometric target. In the final seconds it becomes matter with which the vehicle exchanges energy. High-speed exhaust strikes regolith, turns outward, applies pressure and shear, removes grains, excavates cavities and accelerates particles. Dust can recirculate toward the vehicle, obscure sensors, contaminate mechanisms and settle on nearby equipment. Larger grains can become projectiles. Plume-surface interaction is therefore simultaneously a fluid-dynamics, granular-mechanics, visibility, contamination and site-architecture problem.

Mars already provides direct evidence at robotic scale. After InSight landed, its Instrument Deployment Camera imaged pits excavated by the thrusters. Orbital imagery also showed a darkened zone altered by retrorocket activity. These observations do not tell us directly how deep a 50-tonne lander would excavate; they prove that comparatively small robotic propulsion systems physically modify Martian soil. Ignoring PSI at human scale would be like designing a port without studying what ship propellers and jets do to the dock.

NASA treats PSI as a multiphase discipline. Gas accelerates a distribution of particles with different sizes, shapes and densities. Fine dust follows the flow more readily; larger grains carry more inertia. Low Martian density introduces rarefaction, making continuum assumptions less reliable in some regions. Coupled CFD and granular-flow models therefore attempt to predict erosion, cratering and ejecta while low-pressure chamber experiments provide validation points.

Terminal guidance must decide not only where the feet will touch but how high the engines are, how much thrust they produce, how many nozzles are firing, where they point and how long a patch of ground remains under severe loading. A layout that is excellent for control can be poor for the surface. Moving engines outward or canting jets changes structural loads and moments. Vehicle geometry and landing-site geometry become one architecture.

At low altitude a plume evolves from a relatively free jet into one strongly influenced by the nearby ground. Chamber pressure is far above Martian ambient pressure; exhaust expands through the nozzle, forms shock structures and reaches the surface at high speed and temperature. Part of that flow is turned radially into a wall jet. The near-horizontal jet applies shear stress to the regolith and can mobilize grains well beyond the area directly below the nozzle.

Mars’ low ambient pressure also makes gas rarefaction more important than at Earth sea level. Where molecular mean free path is no longer negligible compared with relevant flow or particle scales, conventional continuum assumptions become less reliable in some regions. This does not mean the entire descent requires molecular simulation. It means qualification must identify where ordinary CFD assumptions stop being trustworthy and where hybrid methods or tests at relevant pressure are needed.

The ground adds another layer. Cohesion, grain size, angularity, cementation, ice, slope and rocks change the erosion threshold. Two zones that look similar in orbital imagery can react differently to the same jet. For a permanent base, geotechnical characterization of the landing area becomes as important as mapping: ground suitable for supporting a habitat is not automatically ground suitable for hundreds of powered descents.

Engine height above the ground is therefore a first-order design variable. Raising engines can allow the plume to spread before impingement and reduce some local concentrations, but can increase structural moments, plumbing mass and integration difficulty. Moving engines outward may protect the centre beneath the vehicle while bringing jets closer to landing legs or nearby equipment. Canting them changes lateral loads again. There is no intrinsically ‘clean’ geometry: every arrangement redistributes pressure, heat, dust and control authority. PSI must be studied while vehicle architecture is still flexible, not discovered after propulsion and structure are frozen.

The problem is also coupled to vehicle attitude control. A small attitude correction changes the direction of one or more plumes and can move the highest ground load laterally. Near touchdown, guidance commands that are trivial from the vehicle’s point of view can therefore change where dust and ejecta are sent. PSI analysis needs representative control histories rather than a perfectly symmetric steady hover, because real vehicles continuously correct position, attitude and vertical speed.

From free jet to wall jet: the physics hidden by a dust cloud

Rocket exhaust leaves a nozzle at a state set by chamber conditions, nozzle geometry and ambient pressure. In the thin Martian atmosphere it expands strongly and can form shock structures before striking the surface. Impingement creates a stagnation region, after which gas turns into a high-speed radial wall jet. That near-surface flow applies shear to grains and can initiate motion. Multiple engines add interacting wall jets and shocks, making the environment three-dimensional and unsteady.

Cratering is not a simple pneumatic shovel. Depending on material, erosion may begin by mobilizing loose grains, then expose a different layer, change the local incidence angle and accelerate or reduce excavation. Cohesive crust, loose sand, cemented soil and blocky ground respond differently. Orbital geology tells a mission a great deal, but not necessarily the mechanical properties of the first centimetres exactly beneath a nozzle. Landing-site design therefore carries geotechnical uncertainty.

Ejecta introduce a separate hazard. Ignoring atmosphere, the ideal ballistic range of a particle launched at speed u and angle θ is R = u² sin(2θ)/g. At θ = 45°, sin(90°)=1. With Martian g ≈ 3.71 m/s², a 30 m/s particle would have ideal range R ≈ 900/3.71 ≈ 243 m; at 100 m/s the ideal value is about 2.70 km. This is intentionally simplified: atmospheric drag strongly affects fine dust, topography matters and actual ejecta have a distribution of speeds. It nevertheless shows why a hazard zone can extend far beyond the lander footprint.

Particle energy matters as much as range. Kinetic energy E = ½mv² also grows with speed squared. A modest fragment at high velocity can damage a radiator, antenna, window, suit or tank. Site models should therefore produce mass-flux, velocity and energy maps rather than a single crater depth. For a settlement, the real question is what impact probability each class of nearby asset can tolerate.

A lifted grain becomes a hazard only once its trajectory is understood. In the simplest drag-free ballistic model, range on level ground is R = u² sin(2θ)/g. On Mars, with g ≈ 3.71 m/s², a fragment launched at 30 m/s and 45° would have a theoretical range of about 900/3.71 ≈ 243 m. At 100 m/s the same formula gives roughly 10,000/3.71 ≈ 2.70 km. These are not predictions of real ejecta; drag, terrain, velocity distributions and grain size alter trajectories. They simply show why high ejection speed can create a hazard zone far larger than the plume footprint.

Risk is not limited to large fragments. Dust can obscure cameras and lidar exactly when the vehicle needs velocity and altitude measurements; grains can contaminate mechanisms, radiators, panels and seals; faster particles can strike nearby equipment. NASA therefore treats obscuration, contamination, energetic ejecta and cratering as plume-surface-interaction hazards. Settlement design must convert those physical phenomena into exclusion distances, equipment orientation and operating rules.

The crater itself can affect vehicle stability. Asymmetric erosion may place a leg near an unstable edge, change engine-ground clearance, or promote re-ingestion and structural impacts. Reusable vehicles also require post-landing inspection of engines and protection. The cost of landing is therefore not only propellant; it includes restoring the site and the vehicle.

Particle size matters because gas and solids do not respond identically. Fine dust can be accelerated rapidly and remain suspended or transported downwind; larger fragments may follow shorter, more ballistic trajectories while carrying far more kinetic energy per particle. A single exclusion radius therefore cannot be inferred from one representative grain. Engineers need distributions: how much material is mobilized, at what sizes and velocities, in which directions, and under which engine settings. Those distributions then become inputs to contamination and impact assessments for the rest of the base.

Wind further separates dust from rocks. Fine particles accelerated by the plume can remain airborne and be transported toward assets that were never inside the ballistic ejecta envelope. Larger particles are less affected by wind but may carry damaging kinetic energy. A base therefore needs both a near-field impact model and a far-field contamination model; protecting a habitat from centimetre-scale fragments does not automatically protect radiators, optics or seals from fine dust.

What Curiosity, Perseverance and InSight teach—and what they do not prove

InSight provides direct evidence of excavation from engines close to the surface. Curiosity and Perseverance embody a different architectural choice: sky crane. Their powered descent stages slowed the system, then held propulsion above the rover while lowering it on bridles. JPL describes Perseverance powered descent beginning near 2.1 km altitude and slowing to about 2.7 km/h by roughly 20 m above the surface before the sky-crane maneuver. This geometry keeps engines away from the science vehicle at contact, but it does not eliminate plumes; descent imagery shows sand moving under exhaust.

The lesson is not that a sky crane scales directly to a human lander. Its geometry, mass and requirements are rover-specific. The broader lesson is that EDL architecture can deliberately separate propulsion from payload contact to manage contamination and interfaces. Future cargo designs could similarly use peripheral engines, a departing propulsion stage or another geometry. Every option moves risk among mass, complexity, stability and PSI.

Recent NASA PSI work is intended to bridge the gap between robotic observation and large-scale prediction. Low-pressure chambers drive jets into simulants; diagnostics visualize flow; Loci/CHEM and granular models attempt to reproduce erosion. Work published in 2024 at Mars-relevant ambient pressure underscores that validation is still active research. A model that reproduces one experiment does not automatically become predictive for a lander an order of magnitude larger; multiple flow regimes and materials have to be bounded.

This caution is essential for vehicles of tens of tonnes. Higher thrust can increase affected area and mobilized mass. Multiple engines can make surface jets converge. Cutoff height, terminal acceleration and lateral translation change how long a patch of regolith experiences high shear. Guidance and control decisions therefore have geotechnical consequences.

InSight provides a concrete reminder. After landing, camera images showed two pits excavated by its thrusters. The scale is robotic and the geometry differs from a future human lander, but the observation proves that rocket jets physically remodel Martian soil. The hazard is not purely theoretical. What remains uncertain is the scaling law for very different vehicle mass, thrust, engine height and terrain.

Curiosity and Perseverance used a sky-crane architecture that kept the powered descent stage above the rover while lowering it on cables. This addresses rover-specific constraints and avoids placing the descent stage itself with engines immediately around the science vehicle. It is not directly transferable to a tens-of-tonnes habitat. The useful lesson is methodological: propulsion geometry and engine height can themselves be design variables for managing terrain interaction.

Mars flight evidence must therefore be connected to ground testing. NASA campaigns use cold-gas jets, low-pressure chambers, imaging and soil-displacement measurements to study relevant environments. Other work instruments or models future systems to measure pressure, erosion and ejecta. The goal is not one perfect test that reproduces Mars, but a chain of models in which each piece is confronted with observable data.

Mission comparison also reveals a time-scale effect. A thruster acting briefly may lift dust without excavating like an engine holding high thrust over one location for several seconds. Throttle history, horizontal translation and attitude-control motion cause the impingement point to sweep across the surface. Models therefore need the time history of thrust rather than a single maximum-thrust snapshot. A heavy lander that translates during final descent may spread erosion over a wider area; holding position may concentrate excavation. Guidance and geotechnics remain coupled all the way to touchdown.

Flight heritage should also be archived quantitatively. Images of pits and darkened zones are valuable, but engineering maturity grows when they are linked to known thrust histories, engine height, local terrain properties and measured vehicle motion. Future Mars landers can carry dedicated PSI sensors precisely so that every landing becomes an experiment. Settlement operations would then build a site-specific database rather than relying forever on extrapolation from unrelated robotic vehicles.

That database should preserve negative evidence as well: cases where little erosion occurred are useful if thrust, height and soil properties are known. They help identify combinations that are benign rather than focusing only on spectacular damage. A mature model needs both high-erosion and low-erosion cases so that conservative limits do not become unnecessarily restrictive for every vehicle and every site.

Designing a real Martian spaceport

A settlement receiving vehicles repeatedly cannot treat each landing as an isolated event in an infinite desert. It needs a spaceport. The first trade is distance. A remote landing zone protects habitats but demands roads, heavy rovers, pipelines or cargo transfer over kilometres. A nearby zone reduces logistics but increases ejecta, overpressure, propulsion-accident and contamination risk. The optimum is an operational compromise rather than a purely geometric one.

The second trade is surface preparation. Compaction, sintering or paving can reduce erosion, but each option consumes energy, machinery, material and maintenance. Pads can crack under thermal cycles and loads; joints can become debris sources. Berms may intercept some ejecta while changing the flow. A sacrificial unpaved zone is simple but must be surveyed after every flight. Landing-pad infrastructure needs its own qualification program much like a terrestrial runway.

The third trade is functional separation. Early cargo landers can arrive far away and deliver the machines that construct later pads. Reusable vehicles may use one dedicated complex while small science landers use another. Propellant handling, batteries, pyrotechnics and crew movement need not share the same corridors. Site planning should consider fire, explosion, short crash, long crash, toxic release, dust transport and emergency access.

TRN has to be coupled to that infrastructure. A safe-target map should mean not merely 'flat terrain' but 'prepared, inspected, clear, structurally adequate and certified on this date.' Local beacons can improve precision, but an arriving vehicle needs alternatives if the pad or beacon is unavailable. The spaceport must be able to say no: bad weather, damaged pad, unexpected obstacle, sensor disagreement or human activity in the exclusion zone should trigger divert or delay.

A mature Martian spaceport can reduce risk by design. A hardened pad limits erosion if its material survives thermal flux and shear. Berms can intercept some ejecta, provided their geometry does not create dangerous recirculation or redirect debris toward the vehicle. Separation distances protect habitats, tanks and arrays. Approach corridors avoid overflight of occupied zones. Every measure, however, consumes mass, energy, robotic time and maintenance.

Local manufacture is attractive. Compacted, sintered or otherwise processed regolith might avoid shipping an entire pavement from Earth. But ‘use the soil’ is not a specification. Mechanical strength, jet erosion, thermal cycling, dust production, cracking, repairability and contamination all need requirements. A process adequate for a slow road may fail under a rocket plume. Pad-material qualification therefore becomes its own discipline, linked to ISRU but governed by propulsion loads.

The first mission still faces the bootstrap problem: no perfect pad exists before the first heavy landing. A realistic sequence can begin with smaller cargo and earthmoving robots, create a coarsely stabilized area, install sensors and beacons, and only then accept larger vehicles. A spaceport grows in layers. This predeployment logic directly connects EDL, autonomous construction, power and settlement strategy.

Pad distance from the settlement cannot be selected by one equation. Greater separation protects people and equipment from ejecta and propulsion accidents but lengthens roads, power distribution, communications, rescue response and cargo transfer. Short separation simplifies logistics while increasing consequences. Design must combine ejecta envelopes, credible propulsion failures, prevailing winds, terrain, rescue capability and future growth. A four-person outpost and a thousand-person settlement cannot indefinitely share the same spaceport geometry. The landing zone must be expandable infrastructure.

At that scale, pad maintenance becomes a standing operational function. After each descent, crews or robots must inspect cracking, vitrified regions, deposits and material loss, then decide whether local repair is sufficient or part of the surface must be rebuilt. Construction robots can do much of this work, but they also consume power, spares and time. A spaceport is therefore never passive ground; it is maintained infrastructure whose availability must match the logistics cadence.

Emergency planning belongs in the geometry as well. A disabled vehicle on the pad can block the nominal arrival zone for months unless there is a second usable area or equipment capable of moving it. Fire, propellant leakage or structural collapse can turn the closest route into an exclusion zone. A settlement-scale spaceport therefore needs alternates and recovery access, not one perfect circle on a map.

Recovery vehicles, firefighting resources and safe access routes must remain usable even when the primary pad is unavailable, otherwise one landing accident can also remove the means to respond to it.

From dust to certification: proving that touchdown is acceptable

Credible PSI qualification starts with observables. Ground pressure, wall shear, erosion depth and rate, ejecta mass, particle-size distribution, particle velocity, cloud opacity, temperature, deposits and sensor loads must be measured or bounded. Without them, 'a lot of dust' remains a visual impression instead of an engineering requirement.

Scaling is the experimental core. An Earth chamber cannot simultaneously reproduce a 50-tonne vehicle, a full plume field, hundreds of metres of terrain and the whole Martian atmosphere. Experiments isolate sub-problems: one nozzle or cluster, relevant ambient pressure, selected simulants and selected stand-off distances. Dimensionless parameters and physical models connect test to flight. Uncertainty in that extrapolation must remain visible all the way to the system decision.

The vehicle can itself become a qualification instrument. Down-looking cameras, pressure sensors, lidar, microphones, particle sensors and engine telemetry can record each touchdown. Protected ground cameras and seismometers can observe from known locations. Early cargo missions can therefore be designed as full-scale experiments that recalibrate exclusion zones before crews rely on the site.

Operational certification of a pad should have a date, configuration and limits: maximum vehicle mass, engine family, thrust and stand-off constraints, weather envelope, surface condition, obstacle state and cleared zones. Inspection after each arrival can maintain, reduce or suspend that capability. A Martian spaceport becomes managed infrastructure rather than a dot on a map.

This closes the specialized EDL chain that began with parachutes and supersonic retropropulsion and continued through terrain-relative navigation. The thread is now visible end to end: remove energy, know the state, select a reachable safe target, then make sure the thrust required to survive does not destroy the place where people intend to live. That coupling is what turns a one-off mission into settlement architecture.

Proving that a site is acceptable requires measuring more than whether the vehicle survived. Instruments can track ground pressure, gas velocity, dust concentration, ejecta size and speed, crater depth, temperature, vibration and visibility. After flight, photogrammetry and lidar can map erosion. As rotations accumulate, the base can compare each event with previous ones and detect ground degradation before it becomes critical.

The numerical model must also be falsifiable. CFD that produces an attractive plume image but predicts no measurable quantity is not enough to certify a site. Researchers instead compare pressure, jet structures, particle motion and erosion shape with tests. The problem is coupled: gas moves the soil, changing soil shape alters the gas, and entrained grains can modify exchange again. Every feedback loop adds uncertainty.

Final certification will therefore combine engine envelopes, geotechnics, local weather, terminal navigation and operating rules. A zone may be safe for one vehicle but not for two closely spaced arrivals; safe in calm conditions but not with wind carrying dust toward the habitat; safe on the first landing but degraded after ten cycles. Treating the plume as infrastructure rather than a fluid-dynamics curiosity is what turns a spectacular landing into a durable spaceport.

Operations also need stop criteria. If inspection finds a pad crack, erosion beyond a limit, a failed pressure sensor or an unusual ejecta pattern, the next rotation should not be authorized by habit. A return-to-service process links measurement, diagnosis, repair and decision authority. That may sound procedural, but it marks the difference between a demonstration and an operating transportation system: an industrial system must know not only how to succeed, but when the preconditions for success are no longer present.

This feedback loop is crucial for a reusable fleet. Measurements from flight N should change models or limits for flight N+1 when needed. If observations repeatedly remain inside margin, uncertainty can narrow; if they diverge, investigation must take precedence over schedule. A settlement far from Earth cannot afford to trade that discipline for cadence, because losing one heavy vehicle can remove people, cargo and an irreplaceable transport capability for an entire launch opportunity.

The same logic applies to weather limits. Wind that is harmless for guidance may still transport dust toward solar arrays, radiators or an air intake. Operations need criteria that combine vehicle controllability with consequences to the base. A landing can be dynamically feasible and operationally unacceptable at the same time; settlement safety is a wider envelope than vehicle survival alone.