The universe is brutal on information. I'm looking for ways to preserve about an exabyte of information for a million years. I'm looking for answers rooted in reality without any lucky circumstances helping it along.

I'd prefer if there were minimal to no systems in place to maintain it.

I'm also hoping the information can be stored at maximum on a planetary scale, but hopefully much smaller.

Any ideas?

It should be relatively easy to extract the information as well.

  • 3
    $\begingroup$ I saw something about memory crystals a while back, could store something like 360 TB/disc. I believe the estimate is that it could last a million years (granted there are no scratches/chips/defects in the disc). $\endgroup$
    – user22049
    Commented Jun 23, 2016 at 19:04
  • 27
    $\begingroup$ "It should be relatively easy to extract the information as well" is a massive deal-breaker. We're sometimes having difficulties extracting information from reel-to-reel tape, which is a trivial storage medium to work with as far as data storage methods are concerned compared even to modern digital storage, and only a few decades old. You want something that is easily readable a million years from now! Also, an exabyte (a billion gigabytes) is an awful lot of data. That said, I suggest you look at the LOCKSS initiative. $\endgroup$
    – user
    Commented Jun 23, 2016 at 19:26
  • 14
    $\begingroup$ Do you really mean "relatively easy to extract the information", or just "relatively easy to read the ones and zeros"? There's a risk that your edit changes the question from, "how do I preserve data?" to "how do I write a message that aliens will understand (and also preserve it)?", because a million years might as well be aliens. $\endgroup$ Commented Jun 24, 2016 at 8:34
  • 30
    $\begingroup$ What kind of information is going to be useful a million years later? Cats may not even exist any more, which will make roughly 40% of today's internet traffic incomprehensible. ;) $\endgroup$
    – TMN
    Commented Jun 24, 2016 at 15:01
  • 3
    $\begingroup$ Languages don't last anywhere near a million years, anyway, which basically breaks your "easy to extract" criteria. See how much 700 year old English you can decipher, for example. Or consider the case of the Rosetta stone. Deciphering a language 1 million years later is just not realistic. $\endgroup$ Commented Jun 24, 2016 at 19:37

18 Answers 18


Use microfiche technology. But instead of using films you use ultrahard plates (made from corundum, cubic boron nitride or synthetic diamond) and use laser technology to etch minimized letters into the surface. Then they are covered with a glass plate which can be easily unscrewed. The glass plate may get blind, but by unscrewing the protected surface can be viewed.

It only needs a microscope viewer to decipher. You can put the parts where the files are and encase them in easily breakable glass. The manual to put them together are visual and easy to understand.

Microfiche store 2-3 Million letters (MByte) onto 30x45 mm which gives us a storage capacity of 1.5 GByte pro square meter. Stack them: If we assume 1 plate is 2 mm thick, we get 75 GByte pro cubic meter. One Exabyte (10^18) means therefore 13 million cubic meters. Storing them 3-dimensional means we can estimate the necessary room by taking the cube root and we get a cube with a length of 250 m.

So in fact that much information can be stored physically. The problem will be that it is not of much use. We do not understand our predecessors which only lived 1000 years ago, old Germanic or Anglosaxon is inscrutable for persons using only modern language.

ADDITION: There were some nice comments below about implementing a Rosetta Stone: Add as much information for deciphering as possible: Manuals with pictures explaining what words mean, telling the same story in different languages, smaller and smaller prints to indicate what the plates contai and use error correction codes to allow information retrieval if the plate is damaged.

What is worrying is the timeframe, we are speaking about geolocial spans when mountains rise and fall. Our standard materials like modern concrete does not last a century, concrete begins to turn back to limestone after 80 years. If we die out now, after 1 Million years the earth will not contain any traces of human activity if observed from space (even the pyramides will have eroded). So the structure must be underground to counteract erosion which will wear down even the most resilient materials in time. It must be in a region with very low geological activity and it must survive what nature throws at it (earthquakes, flash floods). Then it must use materials which are chemically inactive and durable. All in all building the holding place can be as much challenging as creating the plates itself.

Some nice insights from TV tropes here.


  • To mark the storage location, for smaller timeframes (centuries, millenia) insert sturdy posts in lines of cardinal directions where each post can be found not to far apart and where the posts cross the location. For longer timeframes, we can dig out big arrows on all cardinal directions and refill them again. If we do that repeatedly, such landmarks are strikingly visible for extremely long times (meteorite strikes prove the point: While flora and fauna recovered, their marks can be still seen after millions of years).

  • To prevent vandalism by brutes, we can fortify the entrance. With our available alloys we can defeat any attempt to enter the storage room until they have at least mediaeval technology. To enter the storage, the person must solve a mechanical puzzle like the Chinese ring puzzle: Chinese ring puzzle

It is a very simple (does not need maintenance or any power source) and can be made very tough to prevent brute force attacks. If something can solve the puzzle, whatever it is, it wants to know (curious) and it is able to know (intelligent). It also invested time and energy to solve the puzzle, so the entrance is a reward and its contents gains value. While it is still possible that vandalism occurs, it is not very likely.

  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ Commented Jun 29, 2016 at 1:05

Put the data on some sort of storage (doesn't really matter what) and stick it in a spaceship. Send the ship on a slingshot around the event horizon of a black hole so that time dilation means a million years passes for us and only a few years for them.

Now you have a ship, the stored data, and some crew who understand the language and technology of the stored data.

Of course, they don't understand the rest of the universe, but that's another question. :-)

  • 3
    $\begingroup$ This is a really clever idea. $\endgroup$ Commented Jun 24, 2016 at 3:02
  • 17
    $\begingroup$ The environment near the event horizon of a black hole tends to be very hostile. Either it is a really, really big black hole, and hence probably deep the core of a galaxy, or the gravitational gradient is going to be steep and tidal effects non-trivial. $\endgroup$
    – Yakk
    Commented Jun 24, 2016 at 14:41
  • 1
    $\begingroup$ It's unfeasible to dilate time to this extent without crossing the event horizon. $\endgroup$ Commented Jun 24, 2016 at 14:52
  • 6
    $\begingroup$ just send 0.9999999999C speed, 70710 times slow down. $\endgroup$
    – MolbOrg
    Commented Jun 25, 2016 at 0:25
  • $\begingroup$ Make sure you include a few linguists in the crew for the language(s) your information is stored in. The planet's language(s) will not have remained the same in a million years. If you have a few linguists, they can work together to try to communicate with the new inhabitants. A bunch of IT guys who are managing the storage will likely not be helpful in that regard. $\endgroup$ Commented Jun 27, 2016 at 18:21

Use 5-dimensional, ultra high density silica glass discs. In theory, these discs are able to store data forever without reduction in data integrity.

Not only is this possible, it's already been done. You can store ~360 Terabytes of information per disc.

But how do you get 5 dimensions per disc? Pulled from the article linked above:

There are the usually two dimensions (width, height) provided by a piece of silica glass, and depth is provided by writing at three different depths (layers) within the glass. The fourth and fifth dimensions are provided by nanostructuring the surface of the glass, so that it refracts and polarizes light in interesting ways.

Keep in mind that they're not normal glass disks. They're SILICA glass disks. You can put them in a safe made of diamond or whatever material you want if you're just wondering where you put the disks. These discs survive temperatures of 1000 °C. Also, these discs are small - they're ~1 inch in diameter, so they're easy to store; even 2778 of them. This method of storage is not subject to erosion because on normal discs, the data is stored on the surface, but on these ones, the data is stored INSIDE the disc using nanogratings. The glass is also chemically stable too, so it wont react with anything.

So if you want to store an exabyte of information using these, you'll need 2778 discs to store it all.

  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ Commented Jun 27, 2016 at 18:38
  • 4
    $\begingroup$ how do you read it afterward? $\endgroup$
    – njzk2
    Commented Jun 27, 2016 at 20:06

In today's world, let us look at the Rosetta Project.

Imprinted on a disk made of a nickel alloy that contains 13,000 pages of information on human languages. It is 3" across and expected to have a several thousand year lifetime (nickel was chosen for its high melting point and low interactivity with magnetic forces - its hard to warp).

One could easily go to lower densities of data to increase the usable lifetime.

The thing is an exabyte is big. Its about 100,000x the amount of the printed material in the Library of Congress. That's about 23M books. Each book contains on average 64,000 words which is about 250 pages on average. This gives 5.9 billion pages for an exabyte give or take.

For those Rosetta disks, thats actually a conceivable number - half a million disks at 3" in diameter. About 7 in2. This would then give us something that needs on the order of 2250 m2 to store an exabyte at rosetta disk levels. This is 1/3 the size of a soccer field.

Would this last 1,000,000 years? Dunno. Probably not - the disk was more intended for the 2,000 - 10,000 year range. If you wanted to be sure, reduce the data density and start working with issues of geological permanence (consider things like putting it on the moon instead).

A key point with this (compared to other storage mechanisms) is that it is intended to be readable by low tech means. You only need a 1000x microscope. This doesn't depend on data formats or advanced reading mechanisms. Its right there.

  • 1
    $\begingroup$ This. Just make the discs out of something more durable and Bob's your uncle. $\endgroup$ Commented Jun 24, 2016 at 3:36
  • 4
    $\begingroup$ +1 for off-planet storage. If a species has the ability to travel to the moon, chances are good they'll have the ability to find, extract and decode the data. If you just bury it in (e.g.) a salt mine, who knows who will find it and what they might do with it. $\endgroup$
    – TMN
    Commented Jun 24, 2016 at 14:44
  • 9
    $\begingroup$ "Oh, spiffy: Another of these strange, patterned pure ore specimens! Now we can rust-proof our swords." $\endgroup$
    – The Nate
    Commented Jun 24, 2016 at 18:05
  • $\begingroup$ exabyte is big it's less then one GigaByte per each human on earth, rather small then big. Human genome is 3GB, for each person, and even we have similar genes at 1/10 compression ratio will be more then exabyte. If one live will have 1 hour important set of events. Is't much. $\endgroup$
    – MolbOrg
    Commented Jun 25, 2016 at 0:21
  • $\begingroup$ The moon would work, but craft orbiting the earth at maybe 1000km would be more conspicuous and easier to get to. With all these solutions, it's possible that this library will be stored in such a safe place that it will /never/ be found, or if it is, will be misused. So if that's a concern, getting to orbit presents the right technological barrier to make sure they don't get used for fishing weights or something. $\endgroup$
    – sbeam
    Commented Jun 25, 2016 at 21:13

Let's be real, in a million years data storage may be done in the form of hairspray for all we know. How many people still have floppy disk drive? And that was only 20 years ago. The new generation doesn't even know what a floppy is, it's just the save icon now.

Your best bet would be to back it all up on the current format of choice, then keep transferring from format to format as technology goes on. Don't be shy on redundancies. It will require a lot of space. It will require a lot of effort. It will require a lot of maintenance. But you should be able to preserve your data all the way through. You just have to make sure your physical medium doesn't deteriorate before you save it up on the next generation of CD or what have you.

That goes for physical format as well as file format. They probably don't have Word in the future. There's a number of things that can change in the way we treat files in a million years.

Now having the data is one thing, understanding it is another one. Very few languages nowadays rely on drawing cats and squiggly lines, but that's how the Egyptians did it. Say you back up an exabyte of English text. In a million years, people speak the Beta Praxus Prime dialect of Ozatarian and nobody knows what an english is.

You'll have to translate that data into something people will understand, and you'll have to do it at the end. To remain true to the original, I think you should always keep a copy of the original data, and a copy of the translated data. Something may be lost in translation, but you'll still have the original so it doesn't matter too much.

If you plan to translate it multiple times over the course of a million years as you change the format of your backups, you should also keep a trace of that. You'll inevitable end up with a few exaoctets of information instead of just one, although there's a distinct possibility that people in the future will laugh at you if you say an Eo is a lot of data.

Now the last question is: will anybody care about million years old data? It's very possible that the location of the archive will be lost in time, or that the caretakers will decide to do something else with their lives. Or maybe in a couple centuries your country will be removed from history completely for any number of reasons.

You can't predict what will happen in a million years, so it's hard to come up with a solution that will still be applicable then.

  • 3
    $\begingroup$ "8 exaoctets? The screws on my thrall-socket have more storage than that!" $\endgroup$
    – TMN
    Commented Jun 24, 2016 at 13:38
  • 6
    $\begingroup$ "Very few languages nowadays rely on drawing cats and squiggly lines" - can I assume you don't use emoji? $\endgroup$ Commented Jun 24, 2016 at 16:31
  • 1
    $\begingroup$ I know I don't. $\endgroup$
    – The Nate
    Commented Jun 24, 2016 at 18:06
  • 1
    $\begingroup$ one moment you missing. Let say you current capability to record our communications - it's 100% - text, sound, video. Sound and Visual two basic things we using to communicate. That's not all but definitely more then ancient was capable for. Smell taste are byond our reach today, specially as subjective for us, but as chemicals some could be derived from information we have already recorded(chemicals, genetics). $\endgroup$
    – MolbOrg
    Commented Jun 24, 2016 at 23:39

Use satellites.

  • Put a large amount of satellites in space, each with a complete copy of the data to ensure that redundancy is high in case of unforeseen events.

  • Put the satellites beyond geosync orbit, this will ensure neither orbital decay nor collisions with other satellites is a problem. Below geosync orbit there are too many satellites

  • Use a high density storage medium. Someone else suggested 5D ultra high density silica glass discs. These discs are dense enough that putting an exabyte of data in a satellite bound to Geosync orbit could potentially be done with current technology.

  • Include a way to read the data within the satellites or in another place. I don't know what kind of equipment is needed to read these discs, or whatever alternative you may choose. Depending on the complexity you could choose to include the reading devices in the same satellites that hold the data or in another constellation altogether.

  • BONUS: put huge mirrors on the satellites aimed at earth. Iridium Satellites are famous for this, they are a group of comm satellites with very reflective mirrors, their flares regularly seen from earth without the need of a telescope. If you want someone to easily find your data, this could be a good way to do it, after all, it's not precisely natural to see lights regularly shining in the sky, any civilization with the technology to go to space will be aware that the lights are satellites.


Turn the moon into storage

The moon has a surface area of 37.9 million square kilometers. According to WolframAlpha, this is equal to 3.79*10^19 mm^2.

The moon has stayed mostly the same over the past few million years. There is no weathering or erosion, just a few impact craters.

1 exabyte is equal to 8*10^18 bits.

Right off the bat, if you have a scheme that can encode 1 bit in 1mm^2 of surface area (possibly use 2 materials?), you have enough space for an exabyte, plus error correction about four times over.

Let's look at some other ways to increase the amount of data stored (or better yet, the error correction):

  • Make the planet bigger. This may have unintended side-effects (atmosphere, more craters, etc.)
  • Use layers under the surface. However, too far in and you may reach a mantle layer.
  • Use more materials. If each square mm is one out of four materials instead of out of two, you have double the information density.

Your future civilization will know that something is different when they encounter a perfectly tiled moon orbiting a planet.

  • $\begingroup$ Not sure that binary is a good idea (even with ECC). Then you've got effectively two levels to figure out: first, what does each (8/16/32)-bit group represent, then what do the things those groups represent mean? And how are our hypothetical descendants going to figure out which bits represent text and which a picture? I think the problem with binary is it's just too flexible. $\endgroup$
    – TMN
    Commented Jun 24, 2016 at 13:54
  • $\begingroup$ @TMN It can probably be read, especially if you include a guide somewhere (maybe at one of the poles? You have enough room). As a counterpoint, people have been able to decode data that wasn't meant to be read, like the Zimmerman Telegram and the Enigma, so if we are actively trying to make it read, it shouldn't be a problem. It doesn't even have to be binary, just anything that has the equivalent of 1 bit for each mm. $\endgroup$
    – Daniel M.
    Commented Jun 24, 2016 at 14:04
  • $\begingroup$ @DanielM. I dunno, it still seems like you've got two problems: translating a binary string into a symbol in a dead language, then translating the dead language into something intelligible. Basically, you're going to need two Rosetta stones. $\endgroup$
    – TMN
    Commented Jun 24, 2016 at 14:29
  • 2
    $\begingroup$ The Moon sees a constant hammering of micrometeorites, thermal weathering from the day-night cycle, moonquakes from tidal stresses, dust moved by the solar wind, and other hazards. Data stored on the meter scale would be recognizable and probably readable after a million years, but on the millimeter scale, it's likely to be completely obliterated. See, for example, the analysis of Surveyor 3 after only two years on the Moon. $\endgroup$
    – Mark
    Commented Jun 24, 2016 at 19:07
  • 1
    $\begingroup$ @DanielM., since you're working on the meter scale, you're looking at a million materials. $\endgroup$
    – Mark
    Commented Jun 24, 2016 at 19:20

Just brain storming.

1 You might be able to broadcast the information with EM waves. If it's a million light years away, or hits a mirror half a million light years away that might work. Would be cool to be expecting the big encyclopedia galactica dump that's coming in a million years. The expansion of the universe would redshift the photons like crazy though, so they'd have to compensate for that.

2 Maybe you write it on the surface of a blackhole. event horizon of a black hole using the holographic principle.* I don't know how you read the information though. Maybe some kind of hawking radiation read head. You could cop out a little and say "it's not a real black hole, it's an artificial black hole so it has special properties". That could be real lame though.

3 If you were out in the intergalactic space you could probably just store it as bits of matter orbiting some massive center. I don't think much happens out there.

4 Radioactive decay is random nowadays. I don't think we quite know how the weak nuclear force "works" exactly. So maybe you store it as precisely engineered radioactive uranium or plutonium. It'll be silent and not decay at all for a million years, until it starts broadcasting your message over and over. Or maybe you have to turn it on with some weak nuclear force device.

5 You could bioengineer some being that has the code in its DNA. Humans have about 2.5Gigs of DNA so it'd have to be some pretty crazy cells. Life's own maintenance would be the maintenance system I guess.

6 If circling black holes create waves, maybe you can grab bits of spacetime and tie it together in knots. Then you would have this bizarre little space-time hairball but that had information encoded. Maybe something like the South American quipus. Advantage there would be size. The individual knots could be sub planck scale really. I don't know how the knots would stay tight though. Imagine trying to tie knots in the surface of taut spandex. Which I guess you could do if it trapped something hard like a pebble.

7 Whoa it's 23:15 already. Thanks for the brain challenge tonight. Good luck with your story.

  • Update: including the words holographic principle. Wasn't very clear.
  • $\begingroup$ With regards to #2, black holes have no hair. You can encode maybe a few dozen bits of information in the black hole's charge, mass, and angular momentum, and that's it. $\endgroup$
    – Mark
    Commented Jun 24, 2016 at 19:17
  • $\begingroup$ Quite right. I was thinking about the holographic principle, which I totally didn't mention at all. Here's a nice video. Watch it at 1.5 speed. youtu.be/B2ksDczJOAs?t=559 It's a bit of a jump to say we can make a device that can access it all but hey, maybe in a million years, give or take. $\endgroup$
    – DudeFace
    Commented Jun 24, 2016 at 21:47
  • 2
    $\begingroup$ holographic principle you have to add that to your list of ideas - Oort cloud is such hologravic storage and it already contains a record about history of our universe around our system. Not easy to read trough, and write to. But it easy will keep data for 100million years. $\endgroup$
    – MolbOrg
    Commented Jun 25, 2016 at 0:31

Single point of failure will not work. You need to make the data store mobile and self-replicating to give it any credible chance of (a) surviving and (b) being found after 1 million years.

The problem is that if something is easily accessible, you're saying various waves or energetic particles or outside matter can act on it. Otherwise how would you see it, hear it, probe it, feel it, etc?

Over a million years, the fact that the information can be touched will change it. This is (one reason) why arguably the best naturally occurring data system, encoding data in proteins, still isn't able to produce the same information every time. If the first self-replicating "thing" was immune to data entropy, our planet would only have one species (a very simple one).

So you need to make something fairly resilient, but acknowledge that anything that depends solely on resistance to elements and energetic particles and such is impractical. It either creates a single point of failure, or just ends up never being found.

At the the same time you need to overcome the mutation problem. In nature, not every replication comes out perfectly. Your data store will have the same problem since you can't perfectly control all environmental conditions. So you need a "deactivate" signal that you can send to what is believed to be an imperfect replica. And you need a way to judge whether replicas are correct.

So in designing your replicating data store, give it introspection and give it an interface for signaling other data stores of the same type. Make a set in this generation that you are confident have the correct data. Have them inspect their own data and communicate that with the others of the generation. If one is found to have different data, send the deactivate signal to that one. It doesn't need to be destroyed - just deactivated to keep it from reproducing the mutated data. It may end up being the best copy available in the event of a local cataclysm. You might also be able to 'repair' clones that came out with different data, depending on whether you use rewritable storage or not. But read only and write-once storage may be more resistant to mutation, so weigh your options.

For the data clones that reproduce a generation, give them an algorithm to move them away from their peers once they are verified. They only need to stay close enough to the group so that the next generation can be verified by a number of individuals (or else deactivated).

In this way you solve the single-point-of-failure problem and make a passable attempt at slowing down data entropy. It could work over the scale of 1 million years. You may end up with subspecies of your data because of unforeseen circumstances preventing mutations from receiving a "deactivate." But you can potentially deal with that in 1 million years by comparing the subspecies and attempting to reconstruct the original data from them.

For "easy to extract," if I understand that term correctly, I think you need to build an interface in for any intelligent creature to access the data. It's a million years, so technology guarantees and biological guarantees are in the same boat... you don't necessarily know what you're going to be dealing with. So it would be smart to build in several sensory representations - visual, auditory, tactile - as well as some reasonable signaling using radio or... whatever you think creatures will use for long range communication in 1mm years.

So there you go. RepRap 3d printers with an extra exabyte of data payload, you might call them. Best I can do for now, good luck, hope it works for you ;)

  • 1
    $\begingroup$ Welcome to WB. +1 for biology approach and for p2p system verification and for reconstruction and interface. You suggest very complex system. It would be masterpiece. Long living creature, like turtles, less generations less corruption. I imagine them even easy reading just by looking at their shell patterns. Information density is a problem trough. But in general, thats interesting direction, definitely capable for million of years. $\endgroup$
    – MolbOrg
    Commented Jun 25, 2016 at 0:55

Statistically speaking, you'll lose your data, eventually, so the right way to think about it is along the lines of pushing the probability low enough to be acceptable.

As for your storage medium, if you don't want maintenance (e.g. actively making copies) then you can outrule anything that has a measurable decay rate. No magnetic storage, plastics in general, soft stuff like metals (unless in the vacuum of outer space). Instead, think diamonds with holograms etched inside. (Is that even possible? But you get the point.)

Even then, you'll lose your data over and over again on that timeframe; it will get destroyed by random events, so you can't rely on a single copy.

Let's say the a copy has a 50% chance to be destroyed in 1,000 years. If you have a single copy, then you'll lose it with a 99.930709% probability. If you have 1,000 copies, you have a 50% chance of survival (if I'm doing this math correctly; I'm computing the non-destroyance rates for that time period, for the number of copies; would it be this simple?) Anyway, with 10,000 copies, you get a decent 99.9023% survival rate, but you can go higher pretty fast with more and more copies (e.g. 20,000 => 99.99990463%).

You'll also want to use forward error correction to account for partial damage within a single medium. Physical damages usually cluster, so you'll need to make sure every piece of your data is scattered across as much of the medium as possible (holograms already do this, so if that's what you picked you're at luck.) With proper FEC you can get ridiculously robust storage.

You can also use FEC to split up your data across several storage media, if you think your exabyte is too much to make full copies, and if you feel it's feasible to expect your audience far in the future to be able to collect the required number of items. If you choose this way, you may want to include references (e.g. their physical location) to the items as metadata. This way, you'll need slightly more copies, but they can be a fraction of the size each (depending on the numbers in your "M of N" setup.)

As for the medium, I think you'll need to store your data in something 3D because it's hard to achieve the necessary storage density otherwise if you want to stay under the limits of a building...

  • $\begingroup$ I don't think you're using the word "hologram" correctly. A hologram is an image in three dimensions viewable from several different angles. It's a way of displaying data, not storing it. $\endgroup$
    – Grollo
    Commented Jun 27, 2016 at 13:20
  • $\begingroup$ You can use holograms to store data. You just image your data as a matrix of bits, and take a picture of it. The good thing is that you can angle the recording laser light differently to store a huge number of sheets (though with each additional sheet your other ones get slightly blurred; so, there is a limit to this.) $\endgroup$ Commented Jun 27, 2016 at 14:17
  • $\begingroup$ After looking a little closer I found out that this is indeed a thing. The official name is "Holographic data storage", so that's why I didn't find it at first. $\endgroup$
    – Grollo
    Commented Jun 27, 2016 at 14:39
  • $\begingroup$ Yea, it's pretty cool, right? Idk if there are ways already to store holograms on media that's stable over a million years :) But I was focusing on another fundamental aspect of the problem. $\endgroup$ Commented Jun 27, 2016 at 14:49

Use a physical and chemical process that has already demonstrated adequate lifetime.

Bury it in silt in ... maybe not the bottom of the sea, but a physically similar environment in a geologically stable region of the planet's surface.

After 200 million years, an exoskeletal fish can look like this, with every scale visible. enter image description here

(image source)

Scales are black, glossy, and around 2 mm across - this image looks approx life size on my screen, and the fish is compressed to somewhere around 1mm thick - you can see thin black layers in a cross-section of the resulting sedimentary rock, which is very easily worked, as stone goes. It tends to slice into the original layers very easily, with a skilled tap from a hammer, but I imagine a grinding or polishing process to expose each layer may be more reliable.

So a conservative storage system may manage about 1 bit/mm^3 or 1Gbit/m^3, or 1 Exabyte in a mere 8 cubic kilometres, say a 100m deep layer on an 80km^2 lake bed.

I have seen mother-of-pearl sheens on ammonite shells, so the possibility clearly exists for much higher information density (optical wavelength scale), but the sheen often fades on exposure to air, so I wouldn't recommend relying on high density storage.

However, different kinds of creature can coexist in a mixed fossil bed, so there may be scope for storing the same information in different forms - if the high density can be read, it saves a lot of laborious splitting, grinding and polishing to read the mm-scale backup information.


Found a religion

This question exposes many kinds of problems. Most of (very good) answers have centered on the problem of actual preservation of information in a physical medium, and how that medium can be made to last such a long time.

Suppose we have million-year media. Then we are confronted with a group of issues that could be summarized as "Now, how do we preserve the metadata?" By metadata (in case it's not obvious), I mean the following kinds of information about the information in question:

  • Where is it stored?
  • How to read or interpret it?
  • Which parts of the information are arbitrary and which are objective?

That third challenge is a bit subtle. Here's an example of what I mean by it: Suppose you are communicating with aliens using radio transmissions and no human has ever physically been in the presence of one of these aliens. How can you explain which is "left" and which is "right"? While you're chewing on that, I'd like to follow up with "does it even matter?"

Before you comment with all kinds of thoughts on that last paragraph, let's get back to the question at hand. Clearly, there is information about the information that is important. If we try to come up with some way to preserve the metadata along the same lines as the primary information, then we just have moved the problem "down one" to now needing to preserve the meta-metadata.

What's really needed is a system that is self-sustaining, resilient, and flexible enough to be able to operate in the face of disasters, mass migrations, etc. Burying some esoteric materials with important information written in a simple, universal language, with built-in language lessons in a deep chasm on Earth in a place that is geologically stable with a resilient marker of some sort above it will be useless if humanity has to (or chooses to) flee the Earth entirely.

If instead we can motivate a large enough group of people to write down, translate, discuss, pass down, and re-interpret the information for posterity, along with the mandate that the posterity in question continue the process, that might be the best chance to actually have that information available to humanity over an arbitrarily long timeframe.

History shows us that the oldest information we have from our ancestors is almost universally religious in nature. The oldest extant man-made structures (in which much information is preserved) were built by a theocracy.

Now there are some problems with this plan:

  • The nature of religion in society may be changing - i.e., past history is no indicator of future performance
  • Even in just 2000 - 3000 (to say nothing of 5000) years, precise interpretations of religious information have been made questionable by problems in translation and lost cultural references/assumptions
  • Many religions have died out (although we know that because the religious material created by the followers survive the death of the religion)

Regarding the constellation of issues with this plan, which we might summarize as the "telephone game problem" (i.e., when information is passed along, translated, copied, etc., it's almost always corrupted in some way), even with information that has not suffered from the telephone game problem still isn't perfectly preserved. The US Constitution isn't even 250 years old, is written in a language that is the number one or number two most widely spoken one in the world, is still preserved in its original form, and still isn't 100% clearly interpretable, which we know because there are countless intelligent people who can argue over interpretations of various sections of it.

The writers of that information actually built in a preservation system that is meant to be flexible enough to preserve it - the Supreme Court. Once again, we have a dogmatic system that is maintained by the very people who need the information. Such an information preservation system embraces the fact that information that doesn't change format or interpretation actually becomes obscured and lost in time.

Based on where we are now and what we know about science and humanity, if I wanted to preserve information for as long as possible, starting with today, I'd try to get it into the Torah somehow. That book seems to contain the most effectively preserved information in all of human history, and the preservation system (Judaism) has arguably endured greater challenges than any other in history.

  • 1
    $\begingroup$ How big would the Torah be, physically, if it included an exobyte attachment? $\endgroup$
    – JDługosz
    Commented Dec 7, 2016 at 17:24

You probably already know that information is stores in hard discs, solid state drives (SSD's), flash memory etc. as a stream of zeroes and ones.

You could literally sculpt that out on some plains or plateau of a dead world. By dead I mean no geological processes to erode it. The moon is a candidate.

Use a hole/depression for a zero, a cone for a one. Or better yet, sculpt actual zeroes and ones for fun. If each symbol takes one square millimeter, back of napkin calculations* say you could write your data on the moon about five times (might have to flatten it prior to writing, though). You could always make your symbols smaller if you want to.

There are methods to recover lost data from remaining good data in case some meteorite strikes your artwork. Do some googling on "Hamming code". This will require you to write down more than an exabyte, though.

For redundancy, carve it out multiple times on a single planet. Better yet, do this kind of artistic intervention on multiple planets/moons. Charon, Phobos, Deimos, Ceres etc. come to mind.

As for recovering inormation, all you'd need is a sattelite with a good camera. You could read it all in a few passes of the sattelite.

The funniest part of this is that once your civilization is completely gone from this universe and other intelligent lifeforms arise, they may spend eons in fascination trying to figure out all your videos, game saves, emails, porn etc., believing those to be an encoded message with the key to understanding life, the universe and everything. God knows I would put some Rick Astley in there.

By the way, this method is probably good for three orders of magnitude more time than you need, should you wish to troll aliens for a longer span than a million years.

.* An exabyte is 8*10^18 bits. The surface of the Moon is around 3*10^19 square mlilimeters.

  • 4
    $\begingroup$ Nope. The moon has winds and dust and stuff. Plus one could probably physically proof to you that a camera on a satellite wouldn't work to read mm-big structures (not enough light). And there is enough space debris pretty much everywhere flying around and hitting a planet (especially ones without atmosphere to protect them) to damage and hide from view such surface structures a lot unless you make them seriously huge. $\endgroup$
    – Nobody
    Commented Jun 24, 2016 at 1:09
  • $\begingroup$ I added details about it to my answer. Turns out it's still not a very good idea, but for different reasons than I thought. $\endgroup$
    – Nobody
    Commented Jun 24, 2016 at 10:16
  • 1
    $\begingroup$ It needs to be a really dead planet: no star to cause tidal stresses or charged-particle winds, in a region of space without dust particles, and old enough that residual tectonic stresses have been relieved. $\endgroup$
    – Mark
    Commented Jun 24, 2016 at 19:15
  • $\begingroup$ @Mark for million year it will probably survive, recovering will be not easy, although chances to recover 99% of data are probably good. Although 4.53e18 bytes surprisingly not much, even without counting fact it needs to be divided because of duplication for recovery data. $\endgroup$
    – MolbOrg
    Commented Jun 24, 2016 at 23:54

I like the idea of using doped quartz as the recording media. There's going to be all sorts of trade offs between recording speed and size, and the larger the size, the fewer the number of copies. You'd certainly need redundancy. I like the idea of a tomb on the moon, but am not sure that the radiation would be low enough to avoid significant data loss. Clearly, you can't shield from neutrinos, and the radioactive decay processes going on a couple of km below the Moon's surface aren't known by me (if anyone). A couple of tombs on the Moon, a couple of (artificial?) asteroids on highly elliptical (and out of the ecliptic) orbits and that should do it.


I assume it will only be read after that million years, not during it. Because if it's also read during a millions years, then you would be better off just recopying the data to newer storage media and basically take it wherever your society goes.

Bluerays (or CDs, DVDs, etc) should be one of the most long living technologies available right now (assuming you store them properly, detailed below).

(people disagree with me on that. Until I take the time to research the life time of plastics under the storage conditions I described: use a metal as the substrate and recorded layer and some crystal as the cover layer, using the terms from Blueray specs page 2)

They actually have a physical structure which encodes the data, and that structure should be clearly visible as soon as you put it under a good microscope. Of course you can also use fancy cristals like in the other answer. But optical discs are tested, reliable and cheap. Of course if you are just making up a whole world it doesn't really matter. Just use something where the information is stored in the arrangement of huge numbers of atoms which is obvious even if the current method of reading them out is absent (say NAND chips would probably need more examination before it becomes clear how they work plus they are more susceptible to high energy waves flipping bits. Tapes would probably be easier to figure out than NAND chips, but still less obvious than optical discs and could be wiped by magnetic fields because it's not the arrangement of atoms which encodes the data, but the way the electrons circle around the atoms).

(modified) Bluerays for one Exabyte

  • 0.5*10^8 of them at 20GB per disc
  • that is less than 1000 m^3
  • Choose a planet without tectonic activity (on Earth, anything you put anywhere for a long time basically ends up molten in the core if my understanding is correct).
  • The planet needs a temperature around -20 to 50 degrees Celsius (rough estimate), but probably colder would be fine too, just not too hot.
  • Put them in a massive bunker. No doors or anything, just a huge blob of beton and steel.
  • Clean helium (or pretty much any noble gas) atmosphere inside. That also means no water.
  • Your bunker needs to include as much protection from high energy waves as posssible
  • Make the bunkers visible from space (in a way which stays visible for a long time. This is much more difficult than preserving the data)
  • Do multiple separate bunkers each containing the same data, preferably on multiple planets.
  • Reading the data will probably involve destroying the storage unit or at least damaging/opening it in a way which will make sure it doesn't survive long afterwards

Understanding the data

The pure quantity of data should ensure that anyone discovering the discs can figure out the pattern behind it without further instruction (as long as the data on it is fairly varied, say includes at least an uncompressed copy of Wikipedia or similar).

The way that would probably work is that

  • first they would find the codes for letters. Those would appear extremely often.
  • Then they could make out words and start by applying language models and stuff. Considering they would have dictionaries and everything available, with more complicated words explained in terms of easy ones, they should be able to work their way upwards.
  • in parallel they would learn our image formats. They would already know about ways to store images, so they would probably search for them and just apply any image storing schemes they know to a lot of data until they get images they recognize (say of a sun, or of biologic cells, or of probability distributions of electrons in atoms or whatever)
  • by understanding some XML markup and disc structure (which are computer languages, which means they are simple to analyze) they could understand Wikimedia entries in the sense that they have pictures linked to words
  • and so they would have labelled images which they could use to link the basic vocabulary to pictures
  • which in turn would help them understand more of the language so they could understand more of the image descriptions
  • etc

Surface structures

Two answers so far suggested encoding the information on the surface of the moon, in structures where one bit is the size of a square millimeter. There are problems with that. I'll discuss the example of the moon, but should be applicable to other moons/planets.

  • According to Wikipedia 5 tons of particles hit the moon per day. In a million years that's 4.8 mg/cm^2. If the weighted average speed is 20km/s then that would be the equivalent of half a gram of TNT being detonated per cm^2. This is (much) less than I initially expected, but it will still erode most materials quite a lot
  • Because of this you would need to use a relatively think layer of a relatively strong material (say at least 1mm of a metal if you are doing a hole/no hole encoding)
  • You have a huge surface for chemical reactions, your materials would need to be chosen very carefully
  • Roughly guessing from this Wikipedia graphic about the diffraction limit, using a (seriously huge) camera with an aperture diameter of a 100m you could read mm-big structures from up to 200km away. This assumes the surface is illuminated enough, etc., this is a theoretical limit. So basically: it's theoretically possible to read it from orbit, but you would need crazy technology to do it and the 0/1 encodings would need to have a really high contrast (reflects almost nothing / reflects almost everything)
  • using 1mm thick material to coat the moon takes 38km^3 of material
  • you probably can't use material that's already there because it wouldn't be sufficiently strong or wouldn't have the right color or the right chemical properties

So all in all, it's possible, but not practical. If you want something far fetched which may sound fun in a story, fine. If you want something which could actually be done with current technology, better use bunkers containing storage mediums.

The surface would be better spent putting some elements there which are unusual, attracting the attention of scientists. Or maybe unnatural, very large scale patterns (so, many km instead of one mm) designed to be as visible as possible.

  • 4
    $\begingroup$ Plastic won't hold a microstructure on a timescale of thousands of years, much less millions. Storage lifetimes for current optical disks are estimated in centuries at the most. $\endgroup$
    – Mark
    Commented Jun 24, 2016 at 19:11
  • $\begingroup$ Why would it degrade in the absence of oxidants and high energy photons? $\endgroup$
    – Nobody
    Commented Jun 24, 2016 at 19:33
  • $\begingroup$ If you could point me to a source about optical media degrading not because of manufactoring errors, contaminants or uv light... because I couldn't find anything else. $\endgroup$
    – Nobody
    Commented Jun 24, 2016 at 20:04
  • $\begingroup$ because of themperature. This is organic material, and energy of bonds is known. This is polymer, and his properties are defined by length and cross connection of chains. So we may calculate how fast potentially energy will exceed bond energy in polychains them self's, just because they are at room temperature. Same things like UV, just much slower. If you will keep them in 1K temperature, there is radiation factor. Also chemical properties of it may be important. If UV may destroy them it means composition is't stable, potential barrier keeps it together, but single UV enough to change state $\endgroup$
    – MolbOrg
    Commented Jun 25, 2016 at 0:14
  • $\begingroup$ @MolbOrg I should look into that at some point. Qualitatively, I know what you are talking about. I would have thought this shouldn't be a problem (I mentioned the requirement for cool enough temperatures), but I may be wrong. $\endgroup$
    – Nobody
    Commented Jun 27, 2016 at 21:43

This is a very interesting question. Actually, there are some experiments that use plants and their seeds as storage mediums - http://www.dx.doi.org/10.1007/s11248-016-9981-1. For example, citing from http://www.storing-data-into-living-plant.net/experiment :: "1g of DNA can store exabytes of data and this is huge capacity. Protected within a seed or a living plant can last for millenia." Thus, it may solve your problem. However, there are still many open questions...


Here are my thoughts:

the information can be stored as a sequence of numbers, if to this sequence add 0(or two..thousand 0) at the beginning - we get the result of division of something (take the speed of light as a constant) to something else - in our case it will be the speed of the emitted object. To read information we need measured by the speed of our object. The only problem is the accuracy and acceleration measurements.

By analogy we can take as a constant size or weight of the atom and to record information in the form of a ratio of the mass of our facility to ensure that we have taken as a constant.

  • 3
    $\begingroup$ Hi Dmitriy Polovinkin, and welcome to Worldbuilding and Stack Exchange. I don't really see how this answers the question of how to preverse the information; this answer seems to address encoding of information rather than storage and retrieval of information. I recommend that you review How do I write a good answer? in our site's help center. You can edit your answer afterwards such that it better (or more clearly) answers the question as asked. $\endgroup$
    – user
    Commented Jun 26, 2016 at 20:18

Readed all answers and comments. Correction codes are in some. In some are question about structure and decryption that information.


Fractals, or at least principle is very useful. There are 3 images, click will open full image:

enter image description here


enter image description here


enter image description here


All tree images are kinda fractal in informational sense (hmm, if imagine that first one gray-white is made by 1-0)

second is connected to movies data, 3th to biology. In same time, by them selfs, they show image encoding systems-approach, and if someone grasps fact that this is image, he just automatically will get knowledge about approaches we use to encode visual information.

Also interesting moment, faces will be recognized by humans, as long as they are humans. Not sure about million years trough, but assume it's about peoples.

For other species it will be tricky part to recognize that this is information (if we talk about primitive technologies, but even for advanced it may be a challenge, depends how we encode)

Nice thing about such approach, deep leveling information. It may be readable even for stone age peoples. Yes it will take some time for them to read it but, this information may guide them from beginning from stone axes to tunnel microscope.

Each layer is accessible on current level and describes access to next level.

This leveling structure is good also because of density of recording, it may totally consists from subatomic records, but still kinda be accessible from stone age tech (yes, they have to born concept of information as such, but not necessary consciously, sub-consciously will be ok too)

Probably I'm not original, also I would like to see more sophisticated approaches.

this Image is Voyager gold plate enter image description here

As human born in our time and as one who is familiar with concept of plots etc. I have 3% clue what this picture have to mean. Long time I tried not to spoil meaning of that plate, and I think I almost did that(until first smart comment of course - no spoilers pls).

Time will come and I will imagine myself as alien civilization, which tries to guess what hell that means. (almost did that, f spoilers popped up - have to wait next 10y to forget). But those who do not know exact meaning may test them self.

Back to fractal knowledge. Roll out that knowledge is like learning, and this process may be compatible with human behavior, learning style. As small kid learns world, so will be such representation learned. Literary, it may be available almost from begin for a kid, as toy maybe and may lead him trough live to PhD in Physics.

My first experience of English was watching cartoons, although I learned that language barely enough to communicate, but still. But it may be not only vocal language, but also it have not to be mathematics just from start. Because mathematics as whole is also concept, existence of which it is a knowledge to be learned first by it's selfs.

Not sure if it will work on 1kk old humans, eventually it will, probably, specially if form of that storage will be handy to smash something.

There may be much more connections to human nature than just form or visual or even sound (natural frequencies of that storage as example), in data structure, what attracts our attention on refectory level and so on. It may be so specific to humans as whole, so it may be big challenge to other intelligent creatures to read that information without humans, and still be accessible from stone age level for humans.

That way storage may be as dense as it possible for civilization who records it, be cryptic for others, be accessible for those who have to read it, in wide range of their development and knowledge.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .