In order for humans to survive trips to Mars, they’ll need oxygen—both to breathe and to fuel their rockets for travel back to Earth. Now, a team of scientists has designed a robot that could extract oxygen from water on the Red Planet, per a paper published Monday in the journal Nature Synthesis.
The robot, which the team calls an “A.I. chemist,” used a machine learning model to find a compound that could ignite an oxygen-producing chemical reaction on Mars. The compound, known as a catalyst, is made entirely from elements found in Martian meteorites—which means, if such a system could work reliably, space travelers would not need to bring oxygen or even the catalyst needed to produce it.
“If you think about the challenge of going to Mars, you have to work with local materials,” Andy Cooper, a chemist at the University of Liverpool in England who did not contribute to the findings, tells Nature News’ Jonathan O’Callaghan. “So, I can see the logic behind it.”
Mars’ atmosphere contains only trace amounts of oxygen. But scientists have detected evidence of liquid water beneath the planet’s southern ice cap, as well as water ice under the surface. To create the breathable gas, researchers wanted to find a way to break down this water into its hydrogen and oxygen molecules using materials found on Mars.
Their robot analyzed five meteorites that either came from Mars or had a composition similar to that of the Martian surface. Using a laser, it identified significant amounts of iron, nickel, calcium, magnesium, aluminum and manganese in the samples.
From these six elements, the robot’s algorithm determined it could produce more than 3.7 million possible molecules to break down water and form oxygen on the Red Planet. Finding the best one from this extensive list would take roughly 2,000 years of human labor, the study authors write.
But for the A.I. chemist, this deduction process took only a matter of weeks. To determine the best formula, the robot analyzed the data with machine learning algorithms and theoretical models. It made and tested 243 of the possible catalysts, according to Space.com’s Charles Q. Choi.
“On Earth, we don’t use these six elements, because we have more choice,” Yi Luo, a co-author of the study and researcher at the University of Science and Technology of China, tells New Scientist’s Alex Wilkins. “These six elements are not the best for this kind of catalyst, and it limits its performance, but it’s what you’ve got on Mars.”
With its chosen catalyst, the system could produce oxygen from the Martian materials at around minus 37 degrees Celsius, demonstrating the chemical reaction could be feasible on the cold Martian surface.
Notably, the robot was able to carry out the entire process—analyzing the rock samples, identifying the best possible catalyst and producing it—without human intervention.
“We have developed a robotic A.I. system that has a chemistry brain,” Jun Jiang, a co-author of the study also at the University of Science and Technology of China, tells Nature News. “We think our machine can make use of compounds in Martian ores without human guidance.”
Other researchers have already successfully made small amounts of breathable oxygen on Mars’ surface. A lunchbox-sized instrument on NASA’s Perseverance rover, called MOXIE, produced oxygen at the rate of six grams per hour during tests in 2021—akin to a small tree, Vice’s Sarah Wells reported last year. That device works by compressing and heating carbon dioxide from Mars’ atmosphere.
Michael Hecht, lead investigator for the experiment on the Perseverance rover, tells Nature News that it would be much easier to use a scaled-up version of the MOXIE system to produce oxygen on Mars, instead of relying on the A.I. chemist.
Now, the authors of the new study will have to see if their robot can work in the Martian environment, where “the atmospheric composition, air density, humidity, gravity and so on are so different than those on Earth,” Jiang tells Space.com.
“To get [the robot] to work is a non-trivial achievement, because you have to get so many parts working together,” Ross King, who studies the automation of scientific research at the University of Cambridge in England and did not contribute to the new paper, says to New Scientist.