MK ATLAS· Science Atlas

Today in Science

15 October 2026 (Thu)

Four stories today. The colours of the Moon that the Artemis II astronauts saw with their own eyes, flying round its far side last spring, have been gathered in a NASA report; and a new explanation of why the Moon's near and far sides are so different has come from rocks brought back by Chang'e-6. Using AI to widen a reaction map to thousands of routes revealed a hidden shortcut for turning carbon dioxide into fuel; and a discovery credited to AI has raised the question of how the human share should be written down.

Space

On a Moon thought to be only grey, they saw yellow and green

The colours the Artemis II astronauts described aloud as they passed the Moon's far side have been gathered in NASA's preliminary science report.

Last April the four astronauts of Artemis II flew round the far side of the Moon and came home. It was the first time in half a century that people had gone close to the Moon. Six months later NASA has published a report gathering what they saw. The spacecraft's cameras had not been set up for measuring colour scientifically, so in this report the most important evidence about colour was what the astronauts said out loud on the spot.

The Moon usually looks simply grey, but the astronauts spoke of brown, tan, greenish and even bluish tones. Of the Aristarchus Plateau, where old volcanic landforms crowd together, Christina Koch said it looked yellowish, and Reid Wiseman said it was like greenish-brown glass. Scientists think this may be the colour of tiny glass beads thrown out by volcanoes. The brown glass brought back by Apollo 17 was rich in titanium, and the green glass from Apollo 15 was rich in magnesium. The astronauts also saw five flashes on the Moon's dark side where meteoroids struck.

Through the eyes of the Atlas

Galileo first turned a telescope to the sky and saw that the Moon has mountains and valleys. It was the moment the Moon, said to be a smooth heavenly ball, became a rough place like the ground beneath us. Four hundred years later, what human eyes saw up close was not shape but colour, and that colour becomes a clue for choosing where to land next.

A word of cautionThe colours are a record made by human eyes and words, so they can vary with the viewer and the light. They still have to be checked against photographs from orbiting satellites.

Source · Universe Magazine (NASA Artemis II preliminary science report)

Earth

Why the Moon's two faces differ: Earth warmed one side more

Comparing rocks that Chang'e-6 brought from the Moon's far side with rocks from the near side, a team suggests that while the Moon was molten the side facing Earth was hotter, so the magma flowed sideways.

The Moon always shows Earth the same face. That near side is spread with wide, dark plains of basalt, while the far side is bright highland with a thick crust, covered in craters. Why one body should be so different on its two sides has drawn many explanations over decades, but none has convinced everyone.

A team at Nanjing University in China compared rocks that Chang'e-6 brought back, the first ever from the far side, with rocks Apollo and Chang'e-5 brought from the near side. They then offered this picture. While the Moon was still covered in an ocean of molten rock, it had already come to face Earth with one side only. The heat made as Earth's pull kneaded the Moon was stronger on the Earth-facing side, and that difference in temperature set the magma circulating sideways on a large scale. Light rock was carried to the far side and thickened its crust, while the molten material that remained gathered on the near side.

Through the eyes of the Atlas

Newton said the force that makes an apple fall is the same force that holds the Moon. That force does more than pull the Moon: it slowed the Moon's spin until only one face showed, and it kneaded its body little by little. Today's story says the heat left by that kneading may have split the Moon into its two faces.

A word of cautionThe team itself calls this one possible explanation. It is an old question, and comparison with other explanations will go on.

Source · ScienceDaily (Nanjing University · National Science Review)

Chemistry

Widening the map to thousands of routes revealed a shortcut where hydrogen crosses whole

When a team in India used AI to widen the routes by which carbon dioxide becomes methanol on copper from about a hundred and fifty to over nine thousand, calculations that had not matched experiment came into line.

There is a reaction that adds hydrogen to carbon dioxide to make methanol, a liquid fuel. It uses copper as a catalyst, a substance that helps a reaction along without itself being changed. Factories already use this reaction, but exactly which path the atoms take across the copper surface is not fully known. So when it is worked out on a computer, the answer often disagrees with experiment.

A team at the Indian Institute of Science first solved 152 reactions carefully with quantum-mechanical calculations, then taught the results to an AI so it could estimate how hard the remaining reactions are. When the routes were widened this way to 9,389, calculations that on the small map had said the wrong substance would mostly form came into line with experiment. The wide map also showed a hidden route. Hydrogen need not cross over one atom at a time; it can pass to another molecule as a whole hydrogen molecule, two atoms still joined.

Through the eyes of the Atlas

Fritz Haber turned nitrogen from the air into ammonia over an iron catalyst and so made fertiliser. For the hundred years since, chemists have looked step by step at what happens on a catalyst's surface. Today's story is a case where AI estimated more branching paths than people could count, widening the map of that surface all at once.

A word of cautionThis route was found by calculation, and no better catalyst has actually been made from it yet. The large rise in conversion is also a number inside the model.

Source · ScienceDaily (Indian Institute of Science · Nature Communications)

Life

An AI is said to have found a gene-scissors look-alike: how much of it was the people's share?

Over the enzyme set Anthropic said its AI had found, a graduate student in Denmark who has been studying the same enzymes has asked whether his unpublished work was used.

Last month Anthropic announced that its AI, Claude, had searched DNA data for twenty-one hours and found, inside viruses that infect bacteria, a set of enzymes resembling CRISPR. Today in Science covered it on 26 September. Since then a doctoral researcher at the University of Copenhagen has said his team has studied these enzymes for several years and that he once showed Claude manuscripts not yet published, and he has asked whether they were used. Anthropic replied that it does not train its models on conversations with users and that its research team could not see those conversations.

A researcher at the University of Manchester has set out the rules for sharing credit in a case like this. The central enzyme had already been found by a team at another university; what the AI newly recognised was that the enzyme, a repeated stretch of DNA and a possible partner protein form one set. When the same search was run ten more times, it sometimes missed that repeated stretch. When an AI discovery is announced, the share of the people who made the data it stood on should be written down just as clearly.

Through the eyes of the Atlas

Photo 51, taken by Rosalind Franklin, was passed to Watson and Crick without her knowing, and the double helix was solved on top of it. How to share credit is a question science has long suffered from. Today's story shows that the question now arises between people and machines too.

A word of cautionA question has been raised; it has not been shown that anything was used. The discovery itself has not yet been peer reviewed, and what the enzyme actually does is unknown.

Source · The Conversation