Dutch and German boffins have proposed a write-once-read-many storage medium they say should survive for a million years and may be readable after a billion.
Described in a paper titled “Towards Gigayear Storage Using a Silicon-Nitride/Tungsten Based Medium”, available here, the five authors explain that in order to store data for a very, very, long time “a high energy barrier against erasure is required.”
Magnetic media, the paper points out, have a low energy barrier before the state of a bit changes.
“If the system would be kept at a temperature of 0K, there would be no thermal fluctuations and the system would stay in this state indefinitely,” the paper says. Storing disks at zero degrees kelvin is obviously not going to happen, so the boffins invoke “the Arrhenius law” to calculate how often state changes should occur under different circumstances and come up with a state change threshold of “1.8 eV at room temperature” as the value they suggest is needed in a million-year medium.
The boffins then go material hunting and suggest Tungsten as the medium as it has “a high melting temperature and high activation energy, furthermore it has a low thermal expansion coefficent.” which we take to mean it'll be hard to erase without a lot of energy and is so stable it won't change state in ambient temperatures.
Encapsulating the Tungsten in Silicon Nitride will help things along further, because the substance “has a high fracture toughness and low thermal expansion coefficient. Another important feature of the Si3N4 is its transparency to light. A very thin film would also be transparent to electron beams.”
We take that to mean it's also tough and stable stuff, isn't prone to accidental state changes and will allow the Tungsten on which data is written (through a process of creating ridges and peaks for zeroes and ones) to be read by a laser or some other optical gadgetry.
The paper goes on to explain some experiments conducted on the Tungsten/Silicon Nitride medium that show it is possible to write data to the Tungsten. The team also baked the medium for an hour at 473k, about 200 celsius, in an “accelerated ageing test”. The medium remained readable after that test and the authors extrapolate the results to suggest it would remain so at the rather tamer temperature of 300k (about 27 celsius) “for well over 1 million years”.
The boffins admit their work is preliminary and that further and more precise tests will be required before they can declare Tungsten/Silicon Nitride viable for a week, never mind aeons. Similar caveats are offered on just about every study The Reg finds at the intersection of material science and data storage. ®