Microgravity and Orbital Environment

The problem

Low Earth orbit is a free-fall environment in which weight is replaced by a small residual acceleration, and in which a surface is exposed to atomic oxygen, hard vacuum, trapped and galactic ionizing radiation, hypervelocity particle flux and a thermal cycle once per orbit.

A robot in low Earth orbit has no weight to hold it to a surface, so every contact force is also an impulse on the robot’s own base. The base is a vehicle whose residual acceleration is set by crew, equipment and reboost. Outside the pressure shell the same robot sits in vacuum, atomic oxygen, ionizing radiation, a thermal cycle every orbit, and a meteoroid and debris flux, all of which act on mechanisms, lubricants, polymers and electronics.

What has flown

Robonaut 2 ran switch, valve, connector and airflow-meter tasks from a fixed stanchion and was first teleoperated on orbit by a crew member in April 2013, but the report gives no task success rates or crew-time comparison [30]. SPHERES formation acquisition runs in March 2007 reduced pointing error from 150 to 167 deg to under 10 deg in about 60 s in the better runs; ultrasound multipath off module walls, not the algorithm, delayed estimator convergence to 45 to 55 s [29].

Four free-flyers have operated inside the ISS. Int-Ball, a 15 cm sphere of about 1 kg with 12 fans and 3 reaction wheels, ran from 2017 to 2019, motivated by JAXA’s finding that imagery tasks took over 10 percent of astronaut time [33]. CIMON, a 32 cm sphere, held its first session with ESA astronaut Alexander Gerst in the Columbus module in November 2018 [33]. Int-Ball2, a 20 cm sphere of about 3 kg with 8 fans and 3 reaction wheels, made its first on-orbit demonstration on 17 October 2023 [32][33]. Its flight ground truth comes from reference markers with a maximum error of 28 mm and 1.1 deg, so the navigation error reported on orbit includes the marker error [32]. Astrobee replaced cold gas with two impellers and, over six years, logged more than 165 real-time operations and more than 90 Guest Science sessions, with localization error measured on 12 ISS activities [31]; the 2025 joint review counts more than 175 sessions under a different counting convention [33].

Astrobee has also served as a flight testbed for software. AstroLoc2 reduced absolute trajectory error and runtime against AstroLoc on 10 ISS activities, and ran faster than DSO and ORB-SLAM3 at lower accuracy than either [34]. Adding learned object semantics to sparse-feature localization lowered position error in all five ISS datasets, which cover three ISS days months apart in one module [35]. FastCD scene change detection ran on three surveys each by Bumble and Queen in the Japanese module, with no ground truth coordinates for scoring [36]. An Astrobee Wi-Fi survey in 2022 found that free-space inverse-square pathloss does not hold inside ISS modules, with signal strength varying by 20 dB along a single US Lab path and uncalibrated absolute levels [37].

Exposure hardware has flown more often than robots. MISSE 9 through 15 carried 365 polymer and composite samples in ram, wake, zenith and nadir positions; ram atomic oxygen fluence reached 3.93e20 atoms/cm2 on MISSE 10 against 4.46e16 atoms/cm2 on a MISSE 9 wake deck, four orders of magnitude lower [40]. A MISSE 12 ram witness coupon collected 15.1 atomic percent silicon from on-orbit contamination in 0.89 years, and 5 mil aluminized FEP lost 67 percent of its elongation at failure after 0.54 years at zenith [40]. Erosion yields were measured for 150 samples, but the overview withholds them for a later paper. ELSA-d, a 175 kg servicer and 17 kg magnetic-plate client launched mated on 22 March 2021 into a 550 km, 97.5 deg orbit, was designed to demonstrate capture, tumbling-client capture and client inspection; the held paper predates the demonstrations and reports no results [41].

What is being tried now

Free-flyer navigation and perception. Flown on the ISS as software on Astrobee: sequential localization [34], semantic maps [35] and scene change detection [36], each scored against reference trajectories that carry their own error. Ground-verified only: Int-Ball2’s Monte Carlo, hardware-in-the-loop (800 mm arm workspace) and planar granite air-bearing (3.0 x 1.5 m) tests preceded its flight [32].

Learned and optimized control of free-flyers. Proposals tested in simulation only. A reinforcement-learning docking policy for Int-Ball2 trained in Isaac Lab inside a model of Kibo shows that propeller drag torque and spin polarity must be modeled for the policy to work, with no hardware transfer [44]. A mission-autonomy architecture with keep-in corridors and keep-out hulls for a free-flying inspector was exercised in MuJoCo on an idealized rigid body [43]. Propellant-free slew and detumble using a free-floating spacecraft’s own manipulator joints is also simulation only, with flexible links excluded [45].

Low-gravity granular and sampling mechanics. Parabolic flights have validated penetrometry for the MMX sampler: eight actuators, four tip shapes and six materials, with peak reaction force near 30 N in a cohesive Phobos simulant against under 1 N in 2 mm glass beads, where the bed behaves like a liquid in the parabola [39]. On the ISS, a centrifuge held 0.063 to 2.0 G for seven hours, removing the tens-of-seconds ceiling of drop and parabolic facilities; mass flow rates follow Beverloo’s law but deviate at low gravity, and Van der Waals forces dominate for particles under about 200 micrometers even at 1 G [38].

Environment model tooling. Converting NASA ORDEM 3 and MEM 3 flux files to the European STENVI format reproduces risk exactly for ORDEM, but realistic MEM files read through the European tool give substantially higher risk, a bias the paper does not quantify [42].

Servicing and capture. Magnetic capture of a client with a ferromagnetic plate is the approach ELSA-d was built to test [41]; Int-Ball2 uses a magnetic docking station [32].

The numbers

Free-flyers on the ISS

Vehicle	Size and mass	Actuation	Source
Int-Ball	15 cm sphere, about 1 kg	12 fans, 3 reaction wheels	[33]
CIMON	32 cm sphere	propelling fans	[33]
Int-Ball2	20 cm sphere, about 3 kg	8 fans, 3 reaction wheels	[32][33]
Astrobee	not tabulated here	12 nozzles fed by two centrifugal impellers	[31][33]

Reference orbit

Quantity	Value	Source
ISS inclination	51.6 deg	[7]
ISS altitude band, assembly to assembly complete	350 to 450 km	[7]
Orbital translational velocity	7.7 km/s	[4]
Altitude band in which atomic oxygen dominates surface chemistry	180 to 650 km	[5]
Ambient pressure measured on LDEF	1e-4 to 1e-7 torr	[6]

Residual acceleration

Free fall is not zero acceleration. Experiment and life support equipment, crew activity, aerodynamic drag, gravity gradient and vehicle rotation all contribute, and the resulting field is treated as three components separated by frequency [1].

Component	Band	Spatial behavior	Source
Quasi-steady	below 0.01 Hz	Predictable at any point in the vehicle from a single measurement, the center of gravity location and rigid-body kinematics	[1]
Vibratory	0.01 to 300 Hz	Must be measured near the point of interest; propagation through structure is not reliably predictable	[1]
Transient	non-periodic, content across the whole spectrum	Must be measured near the point of interest	[1]

The quasi-steady band is bounded above by the lowest natural structural frequency of the vehicle, and its content varies over periods longer than a minute [1]. Vibratory content begins at that same structural frequency. Orbiter structural modes were identified at 3.5, 4.7, 5.1, 6.2 and 7.4 Hz, and the Spacelab module at 8.5 Hz [2].

Measured disturbance magnitudes:

Event	Magnitude	Duration	Source
ATV initial docking contact, 3 April 2008	12.9 mg	impulsive	[1]
ATV reboost	21 min burn, delta-v just over 4 m/s	21 min	[1]
Thruster firing transient	~20 mg	minutes	[2]
Crew push-off or impact	~50 mg	short	[2]
Payload-induced shock	50 to 55 mg	tens of ms	[2]
Water pump at 10,000 rpm (166.7 Hz)	4 to 5 mg	continuous	[2]
General equipment operation	0.3 to 0.7 mg	continuous	[2]
Crew exercise	band 0.01 to 5 Hz	daily	[2]

Reboost is not optional: the vehicle decelerates continuously under drag and the orbit decays [1]. A manipulator working at a berthing interface therefore does so against a base whose acceleration state includes both a periodic thrust-and-coast history and crew-driven transients at tens of milli-g.

Instrumentation defines what can be resolved. The Space Acceleration Measurement System uses an adjustable anti-alias cutoff between 25 and 400 Hz [1]. SAMS-II remote triaxial sensor heads cover 0.01 Hz to beyond 300 Hz, extending to 400 Hz for transients, store 10 hours of data from five sensors at full bandwidth, and had returned 3.4 TB by Expedition 14 [3].

The measured ISS record over Increments 2 through 4 (May 2001 to April 2002) shows that the relevant number depends on the bandwidth a mechanism responds to [22]. At a 100 Hz cutoff the 95th percentile is about 2.2 mg whether the crew sleeps or not, but filtered to 5 Hz the same sensor reads 0.06 mg during crew sleep against 0.20 mg awake, and docking transients peak near 10 mg [22]. A test that is sensitive below 5 or 25 Hz gains from running during crew sleep; one sensitive to 100 Hz does not.

Ground analogs of free fall

Ground facilities reproduce part of this field and trade quality against duration. The NASA facilities surveyed in 1989 span a 2.2 s drop tower, a 5.18 s evacuated drop at about 1e-6 g, a 100 m tower to 4.27 s, a drop tube at about 1e-4 g for 4.6 s, and parabolic aircraft at about 1e-2 g for up to 20 s; those figures are nominal capabilities, and the KC-135 no longer flies [23]. The Bremen tower gives 4.7 s at 1e-6 g, 9.3 s with its catapult, at three runs per day because evacuating the tube to 10 Pa takes 1.5 to 2 hours, and its GraviTower variant gives 2.5 s at 1e-4 g with 20 runs per hour and a lunar or Martian partial gravity mode at 1e-2 g quality [26]. Payload capsules take about 50 g deceleration, and the guide states quality as an order of magnitude rather than a measured spectrum [26]. On-orbit residuals of roughly 2 mg [22] are about ten times better than the parabolic-flight floor, and no drop facility offers minutes of exposure.

Contact dynamics of free-floating bodies are tested on planar air bearings, which keep two translations and one rotation and give up the other three degrees of freedom. The ESA ORBIT floor is 5 x 9 m epoxy, tracked by twelve Vicon cameras at up to 250 Hz, and the measured residual floor slope sets how long after a contact the drift remains usable [28]. None of these analogs reproduces the atomic oxygen, thermal cycle or radiation exposure of orbit.

Neutral atmosphere and atomic oxygen

Solar ultraviolet below 243 nm exceeds the 5.12 eV dissociation energy of O2, so the residual atmosphere at orbital altitude is predominantly atomic oxygen [5]. At 7.7 km/s each ram-facing surface meets that population at a collision energy of 4.5 eV and a number flux of 5.23e13 atoms/cm2/s at ISS altitude [4]. Erosion is therefore strongly directional: ram surfaces erode, wake surfaces do not.

Ram, wake, zenith and nadir directions relative to the velocity vector. Atomic oxygen erosion is confined to ram-facing surfaces because the 4.5 eV collision energy comes from the vehicle velocity rather than from thermal motion of the gas [5]. Source: [5]. Public domain (NASA).

Erosion yield Ey is the volume removed per incident atom. Flight values from MISSE 2 (3.95 years, 16 August 2001 to 30 July 2005, ram fluence 8.43e21 atoms/cm2) [5]:

Material	Erosion yield (cm3/atom)	Source
Polyoxymethylene	9.14e-24	[5]
Polycarbonate	4.29e-24	[4]
Epoxy, Hysol EA956	4.21e-24	[4]
Mylar, PET	3.01e-24	[4]
Kapton H (reference material)	3.00e-24	[5]
PEEK	2.99e-24	[4]
Kapton HN	2.81e-24	[4]
ECTFE, Halar	1.79e-24	[4]
ETFE, Tefzel	9.61e-25	[4]
FEP Teflon	2.00e-25	[4]
PTFE	1.42e-25	[4]
DC 93-500 silicone	3.81e-27	[5]

At the MISSE 2 fluence, a Kapton H surface loses about 25 micrometers [4]. Silicones survive because atomic oxygen converts the surface to a glassy silicate that passivates the layer below, at the cost of mud-tile cracking of that layer under strain [5]. Kapton H density of 1.4273 g/cm3 is the conversion used between measured mass loss and thickness loss.

The ground facility against orbit, on the same 39 polymers

A radio-frequency plasma asher is the ground facility that stands in for orbit when a material has to be screened before flight, and the MISSE 2 PEACE experiment is the one case where the same 39 polymers were run in both [20]. The asher yield was higher than the space yield for every one of the 39, by a factor between 1.02 and 37.1 [20]. Both halves are measurements: the flight yields are mass loss on vacuum-dehydrated samples of stated density and exposed area at one common fluence, and that fluence, the 8.43e21 atoms/cm2 above, was itself determined from the mass loss of two Kapton H witness samples on the tray rather than taken from an atmospheric model. The asher yields are relative to Kapton H, with the samples in metal holders and the asher operated on air.

The direction is systematic and the size is not. Ground atomic oxygen testing overestimates erosion for every material in this set, but by a material-dependent factor, so the overestimate cannot be divided out: the fluoropolymers sit at 6.1 to 8.0, Kevlar at 24.0 and white Tedlar at 37.1 [20]. Two consequences follow for a designer. An asher screening pass is a conservative bound on erosion and nothing more. And two materials ranked against each other by asher yield can be ranked wrongly, because a 37-fold error and a 1.02-fold error appear in the same table [20]. Kevlar is a woven fabric rather than a film, and the source reads its position near the top of that list as weave geometry being what a ground facility fails to reproduce. That is an inference drawn from the ranking, not a separate experiment.

Mass loss is not a usable erosion metric for silicones at all. A silicone typically does not lose weight in atomic oxygen, because the surface converts to the glassy SiOx layer described above, so a mass-loss ranking places silicones among the most durable materials while that layer is shrinking and cracking to expose the substrate; nanomechanical surface hardness is used instead to set ground-facility equivalent fluences [20]. What changes optically is measurable. An uncovered DC 93-500 silicone sample on the MISSE 2 ram tray went from a solar absorptance of 0.005 before flight to 0.033 to 0.037 after four years [5], a factor of about seven, with total reflectance barely changed and the surface crazed into mud tiles.

Ram fluence scales with mission duration and solar activity: MISSE 4 accumulated 2.15e21 atoms/cm2 in 1.04 years and MISSE 8 accumulated 4.62e21 atoms/cm2 in 2.14 years [5]. On LDEF, inorganic coatings including aluminum, nickel and SiO2 protected T300/934 graphite epoxy from atomic oxygen with negligible mass penalty [6]. MISSE 2 samples also accumulated 1400 equivalent sun hours of solar exposure, and erosion yield for several polymers rose with combined solar exposure and temperature rather than with atomic oxygen fluence alone [5].

Erosion yield of FEP Teflon against accumulated solar exposure in equivalent sun hours, across several flight experiments. Yield rises with solar exposure, so an atomic oxygen fluence alone does not predict thickness loss for this material. Source: [5]. Public domain (NASA).

Thermal cycling

An orbiting structure passes through eclipse roughly every 90 minutes [6]. LDEF accumulated about 34,000 thermal cycles in 5.8 years, over a range of -20 to 160 F (about -29 to 71 C) with a +/-20 F band on those limits [6]. The recorded failure modes were microcracking of composites and delamination of some dielectric and metallic coatings; the aluminized graphite-epoxy noted above survived the full cycle count without coating delamination. Cycling acts synergistically with atomic oxygen, ultraviolet, vacuum and contamination rather than independently.

Heat rejection is radiative because there is no convecting medium. The ISS US On-Orbit Segment rejects heat through the External Active Thermal Control System, two independent mechanically pumped single-phase liquid ammonia loops, each serving three radiators of eight panels, with each radiator carrying two flow paths [16]. Radiator orientation is controlled by the Thermal Radiator Rotary Joint, either parked at a fixed angle or in Autotrack, which adjusts the angle to maximize heat rejection against the moving sink direction. A robot on the outside of such a vehicle has the same constraint: heat generated in an actuator leaves only by conduction to a radiating surface.

Ionizing radiation

Total dose in a 51.6 deg, 400 km orbit originates primarily in the trapped belts, with one solar particle event assumed to occur per mission and add to it [13]. Design limits for the LEO-ISS orbit give a daily integral trapped proton fluence of 1.66e8 protons/cm2 above 0.1 MeV and 5.40e7 protons/cm2 above 0.5 MeV, external to the vehicle [13]. LDEF returned an integrated trapped electron and proton dose of 2.5e5 rad at the vehicle ends against less than 10 rad from cosmic rays over the same 5.8 years [6].

Single event effects, not total dose, set the avionics problem in LEO. ISS screens assembled articles rather than de-lidded parts, in 200 MeV proton beams: protons of 200 to 500 MeV pass through packages, boards and even box enclosures at nearly full energy, and roughly one in 289,000 of them strikes a nucleus inside a device and makes the short-range, high-LET reaction products that cause the upsets [10]. Heavy ion testing runs at linear energy transfer between 60 and 100 MeV cm2/mg at beam energies near 12 MeV/nucleon, and devices sensitive to single event latchup are screened against an LET threshold of 36 MeV cm2/mg [11]. Large solar energetic particle events deliver of order 2.5e7 protons/cm2 above 10 MeV [12].

Trapped-particle models carry stated error against flight dosimetry. Thermoluminescent dosimeters on LDEF read about a factor of two above the AP8MIN proton prediction at depths from 0.5 g/cm2, the measured trailing-to-leading dose ratio was 2.4 against 1.4 predicted, and AE8MIN agreed with thin-shielding doses within about a factor of two, possibly affected by dosimeter saturation [24]. NASA’s 2021 hardness assurance guidance builds on AE8 and AP8, acknowledges the systematic uncertainty from that choice without quantifying it, and contains no part-level test data [27].

Internal, or deep dielectric, charging is a separate mechanism from surface charging. Electrons near 100 keV penetrate 3 mils of aluminum and those near 1 MeV penetrate 80 mils, depositing charge in dielectrics behind the outer shell [9]. The design threshold derived from CRRES data is 2e10 electrons/cm2 accumulated over 10 hours, equivalent to an average flux of 5e5 electrons/cm2/s.

Plasma and surface charging

The ionosphere at ISS altitude has an electron density near 1e11 m-3 and an electron temperature near 990 K, measured simultaneously by the wide and narrow Langmuir probes of the Floating Potential Measurement Unit [8]. Those instruments cover 1e9 to 5e12 m-3 and 500 to 3000 K, with the floating potential probe spanning -180 to +180 V.

Two mechanisms drive the ISS structure negative. The 160 V photovoltaic array strings collect electrons at their exposed high-potential ends, and motion across the geomagnetic field induces a potential of about -10 V on the pressurized truss with a 30 V variation tip to tip across the arrays [7]. Structure potential is clamped by two plasma contactor units emitting below 0.1 A nominally, with a maximum observed discharge current of 100 mA per orbit. With the contactors off, the largest floating potential recorded by the Floating Potential Probe in 2001 was -23 V, at eclipse exit; the effective ion collection area of the ISS structure, the term that sets the electron loss balance, is 35 m2; and above about 160 V, snapover raises the conductivity of exposed array surfaces and increases collection [7].

Charging also bears on dust mitigation hardware. Twelve Electrodynamic Dust Shield panels flew on MISSE-11 in the ISS wake position, two energized at 4 kV and 10 Hz [25]. The glass panel drew current on orbit that matched its ground vacuum behavior, while the Kapton panel showed more than fifteen unexplained current spikes on its first run [25]. The report is preliminary, with two on-orbit runs.

Charging matters for robotics because an EVA suit, a grapple fixture or a manipulator end effector making first contact with a differently charged surface completes a circuit that was previously open.

Meteoroids and orbital debris

The meteoroid environment model MEM 3 covers particle masses from 1e-6 to 10 g, the range able to damage a spacecraft, and gives an average impact speed in low Earth orbit of 19 km/s [14]. Particles below 1e-6 g are assigned a density of 2 g/cm3. Showers contribute about 1 percent of total flux at the 1 microgram limit and less than 10 percent for particles below 0.5 cm, so the sporadic background dominates design [14].

LDEF recorded 34,336 micrometeoroid and debris impacts over 5.8 years [6]; the comparison of MEM against that record works from craters with lip diameters above 1000 microns, a limiting crater diameter of 750 microns [15]. ISS radiators carry MMOD strikes on all panels, and to date no strike has produced a leak in a radiator flow path [16]. Impacts are not uniformly distributed in time; the LDEF record showed them arriving in temporal bursts [6].

Flux depends on orientation as strongly as atomic oxygen does. A manipulator parked in one attitude for months accumulates a different exposure on each face.

Vacuum tribology

At 1e-4 to 1e-7 torr [6] there is no adsorbed water or oxide-replenishing atmosphere on a freshly exposed metal surface, and boundary lubrication must be supplied by a fluid that does not evaporate. Vapor pressure therefore sets the choice of lubricant before viscosity does [17].

Friction against ambient pressure is not monotonic, so a single vacuum figure does not describe a mechanism that sees a range of pressures. AISI 52100 steel sliding on itself gives a friction coefficient of 0.45 in laboratory air at 760 mm Hg, falls to about 0.2 between 1e-1 and 1e-2 mm Hg, and rises again to 0.375 at 5e-7 mm Hg, all measured on a 3/16 in radius hemispherical rider against a 2.5 in rotating disk at 1000 g load, 390 ft/min and room temperature [18]. The rise at the low end is the residual oxygen supply becoming too thin to re-oxidize metal transferred to the rider, and the whole curve sits six to eight orders of magnitude above the roughly 1e-14 mm Hg the same authors give for space.

Below the oxide the contact welds. Matched copper single crystals contacted at 1e-11 torr under a 50 g load with a 10 second dwell give an adhesion coefficient, breakaway force divided by applied load, of 1.02 before any sliding, and after sliding 0.735 cm at 0.001 cm/s the matched (100) pair exceeds 130 against 10.5 for the matched (111) pair [19]. The specimens are 99.999 percent metals cleaned in place by argon ion bombardment and the upper values are the limit of the apparatus rather than measurements, so the set bounds what a plated, machined and handled surface can do rather than predicting it. The ratio between the static and post-sliding values is why a fretting interface is treated as the cold welding risk in an orbital mechanism and a static clamped one is not.

Flight bears that distinction out. A French experiment on LDEF held aluminum, copper, titanium, stainless steel, copper beryllium, silver alloy and palladium washer pairs in static loaded contact at 25 to 135 MPa under Belleville springs for 69 months in low Earth orbit [21], with no sliding motion. Post-flight metallurgical examination of every flight and control pair found no cold welding event, and the ground controls held under vacuum for the same 69 months found none either; the experimenters attribute the null result to the oxide never being removed, because nothing slid [21]. The same report’s review of prior on-orbit experience found no documented significant cold welding failure on a US spacecraft, every reported seizure tracing to vibration-driven fretting or to galling after loss of lubricant. That review is a literature survey rather than a measurement.

Lubricant	Vapor pressure at 150 C (Pa)	Kinematic viscosity at 40 / 100 C (cSt)	Source
PFPAE 815Z	3.5e-6	not stated	[17]
Trisilahydrocarbon 2-94-96	3.1e-5	133 / 20	[17]
Pentasilahydrocarbon MJD990405	1.9e-5	206 / 31	[17]
Multiply alkylated cyclopentane, Pennzane P2001A	2.3e-3	108 / 15	[17]

Extrapolated to 25 C, the trisilahydrocarbon sits at 1.1e-8 Pa and pentasilahydrocarbon MJD991029 at 3.0e-12 Pa [17]. Film thickness under Hertzian contact depends on the pressure-viscosity coefficient, measured at 16 and 11 GPa-1 at 21 and 40 C for 2-94-96 against 11 and 8.5 GPa-1 for Pennzane P2001A.

Wear testing is run in vacuum because the result changes with ambient chemistry. Spiral orbit tribometer runs at below 1.3e-6 Pa, 200 rpm, 23 C, 1.5 GPa mean Hertzian stress and a 50 microgram oil charge per ball take a friction coefficient of 0.28 as the failure criterion [17]. In a reciprocating tribometer at 250 N, 0.7 GPa peak Hertzian stress, 6 Hz and 100 C, a tetrasilahydrocarbon produced a 0.71 mm2 wear scar against 1.02 mm2 for Pennzane 2001A. A bearing test at 1e-5 Pa, 600 rpm, 89 N axial load, 75 C and 20 mg free oil ran about 1000 hours on a pentasilahydrocarbon without measurable degradation.

Lubricated hardware returned from orbit

One set of lubricated hardware has come back. LDEF carried Castrol Braycote 601, a PTFE filled perfluoropolyether grease, on trailing edge drive shafts, ultraviolet exposed with negligible atomic oxygen [21]. The returned grease had a lower base oil viscosity than its control, read as ultraviolet chain scission of the polyether; infrared showed no new carbonyl and therefore no oxidation, but new peaks at 1100 to 1400 wavenumbers suggest the PTFE filler itself degraded, and new DSC endotherms appeared near 106 C and 211 C, the second unexplained. The grease had darkened to black and the cause was not identified [21].

Bonded dry film did worse than either. Everlube 620, a heat-cured MoS2 film in a proprietary organic binder, was gone after the same 69 months on the same trailing edge: only traces of MoS2 remained in the machining grooves, too little to test, the binder having been decomposed by ultraviolet and evaporated [21]. That is the one clear lubricant loss on LDEF and it happened where atomic oxygen fluence was negligible. Tungsten disulfide dry film on the grapple shafts, by contrast, was intact with no discernible difference between ram and trailing surfaces, but no surface analysis and no friction test was run on it, so it is a survival observation and not a tribological result.

A friction number did survive. DuPont Vespel SP-21, graphite filled polyimide, exposed on a trailing edge tray for 69 months, returned a coefficient of friction unchanged from its control at about 0.15 after run-in, over 10,000 cycles against two unflown controls [21]. The friction test was run on the ground in laboratory air, so the 0.15 carries the same room-temperature-air caveat as any bench figure and is not a vacuum value; what the flight establishes is that 69 months of orbital ultraviolet and vacuum did not change it [21].

What is not established

Flight lubricant evidence is one set of LDEF hardware exposed largely to ultraviolet with negligible atomic oxygen [21], so lubricant behavior on a ram-facing mechanism over years is not covered by returned hardware. The ground-to-orbit atomic oxygen factor varies from 1.02 to 37.1 by material [20] and cannot be divided out. The ISS charging model rests on a lumped 35 m2 collection area fitted to data from December 2000 to April 2001, before the probe failed [8]. The MEM validation against LDEF shows a ram-side underprediction that would need a debris share of about 75 percent to reconcile, and no usable debris model exists for 1984 to 1990 to test that [15]. The LDEF dose comparison used 1990s-era trapped-particle models [24]. Microgravity robotics results on ISS are largely narrative: Robonaut 2 reports no success rates [30], and ground analogs omit the non-gravitational environment [28].