Water on the Moon

Most arguments about lunar water are really arguments about vocabulary. Scientists count three things under one name. Hydroxyl is one oxygen atom bonded to one hydrogen, bound into minerals and glass. Molecular water is H2O itself, which can be locked in glass or frozen. Water ice is the solid, which survives only where it is cold enough. The amounts are small. When the airborne SOFIA telescope found molecular water in sunlit ground around Clavius crater in 2020, NASA put the concentration at 100 to 412 parts per million, roughly a 12-ounce bottle spread through a cubic metre of soil, and pointed out that the Sahara holds about a hundred times more. Carle Pieters, who led the Moon Mineralogy Mapper team, said in 2009 that water on the Moon means molecules of water and hydroxyl interacting with rock and dust in the top millimetres of the surface, and that nobody was talking about lakes, oceans or even puddles. So the water lives in three places: at trace levels in the sunlit surface everywhere, inside the Moon in magmas that erupted billions of years ago, and frozen in the permanently shadowed floors of polar craters, the reservoir most mission planners hope to mine. The first two have been measured in samples brought to Earth. The third has been measured at its source once, in 2009, by crashing a rocket stage into it.

For a generation the textbook Moon was dry, and for reasons that looked sound. The Chinese Academy of Sciences' Institute of Physics sums up the era in one line: studies of the Apollo samples found no water-bearing minerals, and a waterless Moon became a basic assumption of lunar science. The traces that did turn up were explained away. Rust in the Apollo 16 breccia 66095, the so-called Rusty Rock, was put down to water carried in by the meteorite that formed it, and hydrogen isotopes close to Earth's were read as contamination. On 7 March 1971 the ion detector left by Apollo 14 recorded a burst of ions that its experimenters reported in 1973 as water vapour ions; eighteen years later two of them concluded that the likeliest source was the lander's own exhaust. The Soviet side came closer. In February 1978 M. V. Akhmanova, B. V. Dement'ev and M. N. Markov of the Vernadsky Institute reported in Geokhimiya that infrared spectra of the Luna 24 drill core from Mare Crisium showed about 0.1 per cent water by mass, rising with depth. They gave the paper a title that ended in a question mark, Water in the regolith of Mare Crisium (Luna-24)?, and were careful: the sample showed little tendency to soak up water from the air, but they would not swear contamination had been avoided. In Arlin Crotts's 2011 history of the subject, no other author had ever cited it. Even the Apollo 17 orange soil, a volcanic glass, was described by NASA in January 1973 as holding no substantial water. Almost forty years later, grains of that same soil helped overturn the verdict.

The polar idea is older than the dry Moon. The Moon's spin axis is barely tilted, so near the poles the Sun always skims the horizon and the floors of deep craters never see it; a 1961 paper by Watson and colleagues argued that water could collect as ice in these cold traps. The polar case began in earnest with Clementine, a small US spacecraft that in 1994 bounced radio waves off the south pole to a receiving dish on Earth. In Science on 29 November 1996 its team reported an echo enhancement confined to permanently shadowed ground, with water ice as the probable explanation; NASA's archive records an estimated deposit of 60,000 to 120,000 cubic metres, a small lake. The claim was soon challenged. In 2006 Campbell and colleagues imaged the south pole with Arecibo radar at 20-metre resolution and found the same signature on sunlit terrain where ice cannot last, following rocky crater walls and ejecta with no link to shadow; if the polar hydrogen was ice, their data were consistent only with grains scattered through the soil. Lunar Prospector counted neutrons instead. Hydrogen slows the neutrons that leak from the surface, and in 1998 the spacecraft found dips in intermediate-energy neutrons of 4.6 per cent over the north pole and 3.0 per cent over the south, consistent with ice buried under up to 40 centimetres of dry soil. NASA announced on 5 March 1998 that water ice might be present at both poles, and NASA's archive puts the ice at about 6 trillion kilograms, an estimate it warns could be off considerably. On 31 July 1999 the spent spacecraft was steered into a shadowed crater near the south pole in the hope of throwing up a plume. Telescopes on Earth saw nothing.

The turn, when it came, started inside the Moon. In July 2008 Alberto Saal, Erik Hauri and colleagues used a newly sensitive ion probe on Apollo volcanic glass beads and modelled how much water the glass had lost as it erupted: a best estimate of 745 parts per million before eruption, and no less than 260. In 2011 they found a way to stop modelling. A Brown University undergraduate, Thomas Weinreich, searched thousands of grains of the Apollo 17 orange soil and found ten with melt inclusions, droplets of magma sealed inside olivine crystals before they could degas. Those droplets held 615 to 1,410 parts per million of water, similar to the primitive basalts of Earth's mid-ocean ridges, and implied that parts of the lunar interior are as wet as Earth's upper mantle. The interior is uneven, though. Basalts returned by China's Chang'e 5 point to a mantle source with at most 1 to 5 micrograms of water per gram for lavas two billion years old, and far side basalts from Chang'e 6 to 1 to 1.5, possibly drier than the near side; the authors suggest the Moon's interior water may be split between hemispheres, as so much of its surface is. How much water the Moon started with remains open, and it matters, because the answer constrains models of the giant impact thought to have made it.

Then, in the autumn of 2009, the surface lit up. On 24 September three papers in Science, from the NASA-built Moon Mineralogy Mapper on India's Chandrayaan-1, from Cassini's 1999 flyby and from the Deep Impact spacecraft, reported an infrared absorption near 3 micrometres across the sunlit Moon, strongest at cool high latitudes: hydroxyl or water, and the data could not say which. A fortnight later, on 9 October, NASA's LCROSS drove a spent Centaur upper stage into the permanently shadowed floor of Cabeus crater near the south pole and flew a shepherding spacecraft through the debris. Colaprete's team counted 155 plus or minus 12 kilograms of water vapour and ice in the instruments' view, and estimated that the soil at the impact site held 5.6 plus or minus 2.9 per cent water ice by mass, along with light hydrocarbons, sulfur compounds and carbon dioxide. The Diviner radiometer on the Lunar Reconnaissance Orbiter put subsurface temperatures at the site near 38 kelvin, and its team read the mix of volatiles as strong evidence of delivery by primitive bodies from the outer solar system. LRO's Russian-built neutron detector, LEND, had chosen the target: it placed Cabeus at 0.5 to 4 per cent ice depending on how much dry soil lies on top, and found that hydrogen-rich patches do not line up neatly with the shadows. In March 2010 NASA's Mini-SAR radar on Chandrayaan-1 reported more than 40 small north polar craters whose interiors scattered radar the way ice does, and NASA estimated at least 600 million tonnes, if the ice was relatively pure and at least a couple of metres thick.

The 3-micrometre band cannot tell hydroxyl from water, and that gap stayed open for eleven years until an aeroplane closed it for one place. On 26 October 2020 Casey Honniball and colleagues reported that SOFIA, a Boeing 747SP carrying a 2.7-metre telescope above most of Earth's water vapour, had caught the 6-micrometre signature that only molecular water produces, at high southern latitudes around Clavius, at 100 to 400 micrograms per gram. To survive the sunshine, they argued, it must be sheltered inside glass or between grains. The same day Hayne and colleagues showed that cold traps exist at every scale from a kilometre down to a centimetre, which brings the permanent cold-trap area for water to about 40,000 square kilometres, 60 per cent of it in the south. Other questions are still argued. Li and Milliken mapped hydroxyl rising with latitude to about 500 to 750 parts per million and changing by some 200 over a lunar day; Bandfield's group, correcting the same instrument's data for heat in a different way, found the signal at every latitude and time of day, with no significant daily migration. LADEE caught 29 releases of water into the Moon's thin exosphere as it crossed meteoroid streams, which Benna's team read as a hydrated layer lying under several centimetres of dry soil, and as a Moon slowly losing water it received long ago or was born with.

China's sample returns have turned the surface question into laboratory chemistry. Chang'e 5 landed in northern Oceanus Procellarum in December 2020 and, before lifting off, measured the ground with its own spectrometer: up to 120 parts per million of water in the soil, mostly implanted by the solar wind, and, if the same model applies to rock, about 180 in a nearby rock. In the returned soil an ion-probe study found grain rims carrying 1,116 to 2,516 parts per million of hydrogen with an isotope signature that points almost wholly to the solar wind, and predicted about 560 parts per million of water in polar soils. In March 2023 He Huicun and Hu Sen of the Institute of Geology and Geophysics reported impact glass beads with up to about 2,000 parts per million of water at their rims, falling toward their cores: solar wind water diffusing inward and, on their model, recharged in under 15 years. Scaled to the whole Moon, the beads could hold up to 270 trillion kilograms, which, as the Academy's release points out, is still far below the roughly billion trillion kilograms in Earth's oceans. In July 2024 Jin Shifeng, Chen Xiaolong and colleagues at the Institute of Physics described a crystal in Chang'e 5 soil, (NH4)MgCl3·6H2O, about 41 per cent water by mass, which the institute called the first discovery of molecular water in lunar soil; the authors excluded contamination and rocket exhaust on chlorine isotope and formation grounds, and the finding rests on one sample. A separate 2024 study showed that melting the soil at temperatures above 1,200 kelvin (about 930 degrees Celsius) yields 51 to 76 milligrams of water per gram by reacting implanted hydrogen with iron oxides. That is water manufactured from the soil. The Academy's release on that study puts the water already in the soil's minerals at 0.0001 to 0.02 per cent.

That leaves the poles, where the stakes are. Some craters sit inside larger shadowed craters and are sheltered even from light reflected off sunlit peaks; their floors reach about 25 kelvin. In 2018 Shuai Li and colleagues found the near-infrared fingerprints of water ice itself in Moon Mineralogy Mapper data, using faint light bounced off crater walls: several thousand 280-metre pixels within 20 degrees of both poles, some possibly around 30 per cent ice mixed with soil, yet only about 3.5 per cent of cold traps showed exposed ice at all. On 18 March 2026 a team led from the University of Hawaii published a new survey from NASA's ShadowCam on Korea's Danuri, a camera 200 times more sensitive than LRO's narrow-angle camera. They found no evidence of widespread surface ice above a detection limit of 20 to 30 per cent by weight and could not rule out widespread ice at lower concentrations; a few spots 20 to 50 metres across could hold more than 10 per cent ice, though other explanations remain possible. In May 2026 Rishitosh Sinha and colleagues, using the dual-frequency radar on Chandrayaan-2, reported signatures of subsurface ice beneath four of nine doubly shadowed craters inside Faustini, Haworth and Shoemaker. Ice may not need permanent shadow either. Chandrayaan-3's thermal probe measured 355 kelvin on a sunward slope of 6 degrees at its high-latitude landing site, hotter than expected, and the modelling built on it suggests that poleward-facing slopes steeper than 14 degrees at high latitudes could harbour water ice, in places that are technically less challenging to reach.

So how much is there? What has been measured is modest: parts per million in the sunlit surface and in returned samples, up to about 2,000 in the rims of individual glass beads, and one polar crater, Cabeus, at about 5.6 per cent at a single impact point. Everything larger is an estimate, and the estimates count different things. The Lunar Prospector era figure was about 6 billion tonnes for both poles. Mini-SAR's was at least 600 million tonnes for the north alone. A 2022 synthesis by Brown and colleagues put roughly 34 million tonnes in the eight south polar cold traps most likely to be rich in volatiles, and a 2007 model gave Shackleton crater 1.6 to 4.5 million tonnes, depending on assumptions. At the other end, Cannon and colleagues modelled billions of years of ice delivery interrupted by impacts that bury it, and found that cold traps older than about 4 billion years could hide buried, ice-rich deposits on the scale of a billion tonnes, while conceding that ice buried that deep is hard to confirm. On 14 September 2026 Martin Elvis and Jonathan McDowell took one billion tonnes as a generous baseline and calculated that it would supply a city of a million people, recycling water as well as the International Space Station does, for only about a century. The origin is just as unsettled. The Diviner team read the mix of volatiles in the LCROSS plume as pointing to comets and other primitive bodies, the isotopes in Chang'e 5 grains point to the solar wind, and the Apollo melt inclusions show that lunar magmas carried water their eruptions could release; Erik Hauri, an author of the 2011 study, suggested some polar ice could have come from that source. How much each contributed is unknown.

That uncertainty is the reason for the traffic. Water means drinking water and breathable oxygen, and split into hydrogen and oxygen it becomes rocket propellant, so ice dug up locally would spare a base from hauling all of it from Earth. The people building the landers want ground truth. Tang Yuhua, deputy chief designer of Chang'e 7, told China National Radio in February 2025, in a report carried by Xinhua, that remote sensing had only suggested ice might be there, and that the mission would go to the surface to verify whether it exists; its hopper is to carry a water-molecule analyser and is designed to fly from sunlight into shadowed craters to fix where the ice is, how much there is and how it is spread. On 23 August 2026 China's crewed space office said Chang'e 7 could not fly in this year's scheduled window, and gave no new date. NASA cancelled its ice-prospecting rover VIPER in July 2024, then in September 2025 gave Blue Origin an option to land it on a second Blue Moon Mark 1 by late 2027, to be decided after the first Mark 1 has flown. After a New Glenn rocket exploded on its pad in May 2026, Blue Origin said in August that it expected that first launch in early 2027. IM-2 Athena reached the south polar region on 6 March 2025 with a NASA ice drill but came down on its side, and the drill could demonstrate only its range of motion. The Japanese and Indian LUPEX rover is designed to drill for polar water and measure its quantity and quality. And NASA, which on 27 February 2026 made Artemis IV its first crewed landing, targeted for early 2028, said then that it aims to send astronauts to the lunar south pole in 2028 and expects to begin building a Moon base on Artemis V, planned for late 2028. Who may own the ice is unsettled too. The Outer Space Treaty bars national appropriation of the Moon; the Artemis Accords, signed by 73 countries as of 21 September 2026, affirm that extracting resources does not in itself amount to appropriation; and a United Nations working group is still drafting principles for space resource activities, with meetings scheduled into 2027.