# Would a frozen Earth "brick" abandoned datacenters?

I am writing fiction about Earth becoming a rogue planet. I'm thinking about how future archeologists would deal with 21st century history. Datacenters might be treasure troves for them, as long as the data can still be retrieved.

What I don't know is wheter a hard disc or flash memory unit can be destroyed by cold, directly or indirectly - or if the data in it may be erased.

Can computers be destroyed, or the data in them be wiped out just by being exposed to an arbitrarily low temperature? If so, what would that temperature be?

Indirect destruction is ok - I remember that in the book The Martian, the protagonist bricked a laptop by trying to use it in the atmosphere of Mars - the LCD instantly boiled due to the low pressure (I consider that a loss because he didn't have the resources to fix it). But here I'm thinking more about temperature.

(Note: bricking)

• What sort of timespan are we talking about here? Dec 17 '20 at 5:00
• How cold are we talking? Once the atmosphere liquefies, that would probably cause problems. Quite aside from most data storage not being rated for vacuum, the lack of atmosphere would expose a lot of unhardened hardware to cosmic radiation. Dec 17 '20 at 5:04
• Few millennia is hugely problematic even before we get to the Earth freezing. Dec 17 '20 at 17:24
• Our current technology is absolutely unsuited for independent recovery. I realize that both the question and the answers are geared towards recovering the bits which make up the data - but when you have the bits, you are nowhere near in decoding what they are saying! You will encounter the already mentioned "encryption problem" on every step of the way, in layers upon layers of random engineering decisions. Our hardware can only read our data because it implements a stack of convoluted protocols unknown to any other civilization. Dec 17 '20 at 18:26
• Indeed, if you allow access to printed books on the needed standards, it becomes much more doable. Dec 17 '20 at 18:35

# Sorry, magnetic data does not last.

Flash memory degrades over time. The relevant industry specification JEDEC JESD218A endurance specification requires a flash memory chip that’s turned off and stored at 25˚C to retain data for 101 weeks. Not even two years.

In practice, you start encountering single-bit failures within 5 years, and complete storage collapse within 20 years. It'll be quicker if you live at high altitude, or if the storage location is not air-conditioned and thus experiences temperature swings.

Mechanical Hard Disks drives fare a bit better. While most of the mechanical aspects of the drives fail quite quickly (2-7 years) when inactive (an unpowered hard drive will fail sooner than one in continuous powered use), the actual data on the platters will last a bit longer.
But the magnetic domains on the storage will creep, fade, and decay in a process known as "bit rot". This failure rate will vary a lot, depending on environmental factors and the manufacturer of the media. Hot is bad; cool but non-freezing temperature is better. Ice formation is very bad. Vibration is bad. Radiation is very, very bad, even the everyday normal background- and cosmic radiation.
A drive in an underground data center that was originally climate controlled, with effectively moisture-sealed air and with significant radiation protection from cosmic rays due to its location and no Radon source(not dug into granite), will be about the best possible location. Yet even there, the media will have some errors within 10 years, many errors after 20 years, and apparently blank storage platters after 100 years.

Magnetic and electrical storage simply do not last -- they are optimized for low-cost, high-speed reliability in the short term only.

Other media
Just about the only bulk data storage that we normally use that will survive on a decades-to-centuries timescale is printed optical disks. Note not home-written, dye-based optical disks, those are worse than all the others here, lasting less than 5 years! We are talking mass-produced, stamped-in-metal-foil-on-polycarbonate-disk CDs and DVDs. With even a modicum of protection against the elements, a good quality DVD should retain its data for millennia or longer. They will become brittle and delicate, but a good archeologist is not bothered by such things.

Note that we can build a magnetic hard drive to retain its data indefinitely. At least on the century timescale. The process involves forming the platter with distinct "islands" of magnetic material, instead of the continuous magnetic domain that is normally used. But NO current commercial drives work like this, because the manufacturing cost is immensely higher, the maximum storage density is a magnitude lower than for normal drives, and the read rate of the drive is also correspondingly lower.
But it would be possible to build a magnetic hard drive that will last "forever", for such purpose as a Time Capsule, Data Ark, ultimate backup of important military data, etc.
But it would need to be planned and manufactured for just that purpose. Consumer-level data centers would not be using such tech.

There are also storage methods that will last for centuries, even millennia. It is not so hard to store data indefinitely, all that is needed is a storage that focuses on long-term safe storage, not ease of access, speed of access, or minimizing of cost. (all three of these are prime drivers for datacenters, of course!)
Example: the microfiches stored in a mine, suitably inerted and stored paper books, some forms of DVD, etc.
There will definitely be huge amounts of data available to archaeologists, but common usage datacenters will not be a part of that.

• Comments are not for extended discussion; this conversation has been moved to chat. Dec 19 '20 at 1:15

Archaeologists probably are interested in data 500 years old and older.

The German government is currently storing one barrel of microfiche per year, containing what they perceive as relevant information, in an unused mine. Those microfiches are made to last and to be readable by low tech. They are guaranteed for 500 years and plausibly readable after maybe triple that time. A real treasure for archaeologists!

In Southampton, a company developed a glass storage disc that may outlast our civilization. The data stays readable for 60 million years, if you treat the disc gently. Sadly I have never heard from them again after they made their announcement. Your archaeologists would need quite high-tech laser readers to decipher those, though. I'll leave a link to the announcment for you if you're interested; and a second link to the advertisment site of the company.

• This is good to know, thank you! Dec 17 '20 at 18:02
• @TheSquare-CubeLaw You should note that random data centres are unlikely to survive long in an apocalypse, but I bet there's at least a dozen groups out there dedicated to storing eg all the text of wikipedia in a massively durable medium.
– Kaz
Dec 17 '20 at 18:04
• "one barrel of microfiche" - that's not a SI unit, right? :) Dec 17 '20 at 19:24
• In 2019, Microsoft recorded Superman on some kind of quartz glass. So if that ends up becoming a go-to method of data archiving, you then could easily last hundreds or thousands of years. Maybe longer. And one note in an article I found directly mentioned the University of Southampton. Quote from article "a virtually unlimited lifetime at room temperature." No mention how the OPs arbitrarily low temperature would treat them. Dec 17 '20 at 22:43
• @John Dvorak they literally use metal barrels like you can see elsewhere for transporting liquids. Here they are meant to keep out the liquids. It was meant to be a plan B if the nuclear war was coming and going, for re-kickstarting everyone into a civilization. I think the Germans are not the only ones who do this. Alexander, yes it is a SI unit if the barrel has 1000 litres :-) Dec 18 '20 at 13:21

While low temperatures are beneficial to integrated circuits because they inherently slow down diffusion processes which on the long run would tamper with the precise arrangement of the circuital elements and all its stacks, they are still a source of damage for the circuit.

In particular with temperature going down, any material will shrink at a certain rate. If the contraction rates of adjacent material is too different and there are no other means of releasing the built up stress, some material may crack. The lower the temperature drops, the higher the chance that a crack occurs. This is the precise reason why deep space probes need a thermal source to stay warm.

I doubt that a data center electronic would be designed keeping in mind operating at the temperatures of a rogue planet, thus the scenario I described above is very likely to happen.

• Can the archeologists fix the crack? or read the data around the crack? Dec 17 '20 at 17:16
• @TheSquare-CubeLaw That PopSci article is completely wrong about how difficult it would be to read data from a shattered hard disk platter - it is is a far, far more daunting task than taping bits of paper back together (platters are mostly made of glass nowadays, I imagine you understand how that behaves when shattered). Assuming you can physically reconstruct the platter to the point it can be read, the shattering will have disrupted the extremely sensitive magnetic fields storing the data, so you are likely to get partial data at best. Dec 18 '20 at 16:57
• @IanKemp I understand that it is hard, and will not give good results, but in the end it is possible; This is kinda like current archeologists dealing with old manuscripts (which have been burned, torn apart etc.), I figure. Dec 18 '20 at 17:11
• @TheSq: except that old manuscripts have data in single characters, not in blocks of data + ECC codes of 4KiB of data encoded somehow (a hardware sector on rotational media). I'm not sure what kind of angular distance that is around the platter, but you might need a significant portion to know whether the data you read means anything (by having a whole sector you can decode). Letters on paper are generally huge compared to the resolution of the ink + paper, i.e. lots of difference between an a and an o, but high-density digital media tries to approach that limit with error correction. Dec 20 '20 at 0:16
• (I don't know if data on a platter is typically encoded as "the data" and then some fully separate forward error correction, or if the whole 4k of data is distributed throughout whatever signals are on disk. It's likely that there's some kind of scrambling / encoding so long runs of 0s in the data don't lead to long runs of the same domain on the platter, but this encoding period might be shorter than the error-correction block size, i.e. HW sector size.) Dec 20 '20 at 0:18

There are already very good answers on this question, and I'd like to point out something a bit out of scope from the question, which I believe is relevant.

There is a lot of data that, if you don't know the format of, is indistinguishable from noise. So even if the data survives on the DVD, drive, disk, microfilm or whatever storage medium you end up using, the format of it will be most important to would-be archaeologists. And that is ignoring the technology necessary to read said data in the first place.

Try reading a word document without word installed and you might get an inkling of how hard it might become. You might be able to hand-wave it away though...

• +1. For .doc files it would be very hard. .docx files are XML, which makes it easier, but they are also zipped which requires knowing the zip algorhitm in order to extract the info. Dec 18 '20 at 9:33
• Archaeologists make it a profession to find out how to decode material where all instructions how to interpret it have been lost. It's somewhat akin to deciphering hieroglyphs, which was a hard problem but ultimately got solved. If a future civilisation found old .docx files, it would require several person-decades of work to fully decipher them, but they know the concepts of container files and zip compression, they will likely be able to decode that stuff - not completely, but enough to retrieve most information. Dec 18 '20 at 13:33
• @toolforger Hieroglyphics were not deciphered by hard work, we simply got lucky by finding the Rosetta Stone which was enough of a key to give us the context to understand and extrapolate. Without that key we would almost certainly still have no idea of the meanings of any of them. Dec 18 '20 at 17:01
• The Rosetta Stone was just one (important) stepping stone, and it still required a lot of hard work before (to make the Stone even useful) and after (to fill all the gaps). No Rosetta Stone existed for Cuneiform, yet it was deciphered. Of course, if future archaeologists don't know about ASCII or Unicode, they will have a really hard time deciphering anything, but given enough time they should be able to infer document structure ("works" vs. "does not work") and compression format (low-entropy result). They will still not understand words like "Trumpism" or "Obamacare"... Dec 18 '20 at 19:32

Magnetic media, look at the other answers. Flash storage is even worse.

There are however writable DVD's generally sold as mdisc that manufacturers claim will last 1,000 years, due to the use of inorganic storage layers. They are read compatible with DVD devices.

Inorganic blu ray media is similar to mdisc and will probably also be legible.

The trouble then becomes finding an operable reader.

Since optical storage is used by cloud providers for lesser used data, and blu ray is generally inorganic storage, more may be readable than you think, subject to finding a working drive. Organic dye based dvd-r is likely to rot as per the other answers.

• Those DVD would last a thousand years if they were produced the right way. Sadly they use a plastic disc, put a metal foil in it, press the information, and then instead of protecting it from the back side, they apply a thin layer of paint on the metal foil. DVD are near invulnerable from the reading side, but from the colourful printed side, they are incredible fragile. They don't survive 1000 years in practice. Sad thing is, they could if produced the right way. Dec 20 '20 at 13:31
• @Anderas do you have anything to back this up? what you say is true of CD's where the data layer is right at the back of the disc, they wanted the read beam to be better able to focus past contaminants. However with dvds the dava layer is in the middle of the disc, there is no way that the printed side will be weaker. Dec 21 '20 at 10:29

The atmosphere will freeze solid. Mainly over the oceans. It will be liquid for a while. The process will produce rather harsh climate that will collapse most buildings.

And don't even get me started on how brittle everything becomes at -40C. Below everything is worse. Uneven thermal shrink + already brittle material = every single chip or magnetic record is gone.

Sorry, but no.

Paper may survive better, if exposed only to vacuum + cold and not to strong winds. Better if a storage is prepared beforehand, in a mountainous region (in order to avoid flooding with liquid nitrogen) and deep underground (to use some geothermal heat).

• It would take a long time for liquid-air temperatures, wouldn't it? The earth's core has a lot of thermal mass, and a bit of radioactivity, and that might keep the ground (and underground indoor areas) from getting that cold for thousands, maybe millions of years. Total wild guess on the timescale. Surface temperatures would asymptotically approach equilibrium with the background of space (~2.7K) over time as the earth cooled, but the question is where along that equilibration process the archaeologists find the Earth. Dec 20 '20 at 0:21
• Core will be white-hot for another few billion years, sure. The surface is another matter. Good timescale estimations here en.wikipedia.org/wiki/Nuclear_winter Dec 20 '20 at 9:27
• Which part of that large article say anything about temp changes anywhere near as large as what you're suggesting? Or heat from the core, or anything like a timescale for liquefying the atmosphere? (That couldn't happen while still in orbit around the sun, even with very dark skies, I think, so which part of the article did you mean?) The larger the temp gradient between air and core, the more heat will flow. (So it may be difficult for the air to cool the surface dirt to temps that would liquefy air indoors in underground facilities.) Dec 20 '20 at 18:32
• @PeterCordes Right now, we have ~170K places (deep cracks in the ice, Antarctica). en.wikipedia.org/wiki/Lowest_temperature_recorded_on_Earth ... and Antarctica is in fact very rich geothermally. The heat we enjoy is from the Sun. The Earth core supplements, on average, single digit percent. At best. Dec 20 '20 at 19:28
• So heat from the core might be enough to keep hydrogen or helium from liquefying for a long time, but maybe not other gasses whose boiling points are a lot farther from absolute zero (or from the background temp of space) Dec 20 '20 at 19:36