The work could open up an entirely new branch of chemistry, as it marks the first time that purely mechanical energy - the stirring - has been shown to produce a chemical reaction.
If it can be improved, the method could be used to beat the greenhouse effect - by generating hydrogen to fuel "water-powered" cars that would produce no pollution.
A team at the Tokyo Institute of Technology led by Professor Kazunari Domen discovered that when they put a simple catalyst into a beaker of water at room temperature, and stirred it energetically, hydrogen and oxygen were given off.
The process even worked in the dark - suggesting that it was the stirring, not any energy from light, which was helping the process. The suggestion stunned Western chemists.
"It's completely contrary to thermodynamic principles," said Professor David King of the chemistry department at the University of Cambridge. But he could find no fault in the scientists' work. But using the discovery to beat the greenhouse effect would need a solar-powered, rather than a mechanical solution.
"The problem is that you need energy in the first place in order to split the water," said Professor King.
"That means that you end up putting more energy into producing the hydrogen than you get out of it."
Professor Domen is collaborating with the Nikon Corporation of Japan, aiming to produce a solar-powered system for splitting water to produce hydrogen.
The gas could then be stored in "fuel cells" on cars of the future: the cells would combine the hydrogen with oxygen to produce an electric current for a motor.
The suggestion that just stirring water in the dark could produce hydrogen gas is so bizarre that to some chemists it recalls the "cold fusion" debacle of 1989, when two scientists in the US claimed to have made atoms of hydrogen fuse in a laboratory test tube.
Subsequent research suggested the experiments were badly flawed.
But unlike that work, Professor Domen's discovery was examined by other chemists before being published in a learned journal, the Royal Society of Chemistry's Chemical Communications.
Professor Domen admits he does not know how the reaction works - only that it does. He reckons that less than 5 per cent of the mechanical energy from stirring is turned towards splitting the water. Commercial exploitation would require a figure at least 10 times better.