back to article Q: How does hydrogen turn into a metal? A: Hang on a second, I need to train my AI supercomputer first

Scientists have trained a neural network on a supercomputer to simulate how hydrogen turns into a metal, an experiment impossible to reproduce physically on Earth. Under extreme pressures and high enough temperatures – such as in the cores of Jupiter, Saturn, Uranus, and Neptune – hydrogen enters a strange phase. The electrons …

  1. UCAP

    Small niggle

    Under extreme pressures and high enough temperatures – such as in the cores of Jupiter, Saturn, Uranus, and Neptune – hydrogen enters a strange phase

    Small niggle here; Uranus and Neptune are too small to have metallic hydrogen in their cores - the pressures are simply not high enough. Saturn is probably large enough for a small metallic hydrogen core, although some models say otherwise.

    It should also be mentioned that metallic hydrogen has been created on Earth using a diamond anvil, albeit for only a few milliseconds. However the continuous state change from a non-metallic to a metallic state has already been observed.

  2. Mike 137 Silver badge

    Another possible niggle

    High pressure and high temperature are typically mutually antagonistic to maintaining the solid state of matter. Is this really a "metal" state as normally interpreted, or is "metal" being used more loosely on the basis of the conductivity phenomenon?

    1. MiguelC Silver badge

      Re: Another possible niggle

      Mercury, gallium and indium are all liquid at NTP, and all are metals too. In physics, being a metal is more commonly regarded as being electrically conductive. Lots of elements not classified as metals become metallic under high pressures and temperatures, while other lose their metallic properties under those conditions (like sodium)

      1. UCAP

        Re: Another possible niggle

        Water is also a solid (i.e. ice) at high temperatures so long as the pressure is also high enough. IN fact some of the water ice types at high pressure/temperature combinations are pretty interesting!

        1. Symon Silver badge

          Re: Another possible niggle

          "pretty interesting" You mean like ice-nine?

        2. Glenturret Single Malt

          Re: Another possible niggle

          Water isn't an element.

      2. Symon Silver badge

        Re: Another possible niggle

        Both gallium and indium are solid at NTP. Gallium melts at 30C, which is close to the 20C of NTP. Indium melts at 157C, so God knows what you're getting confused with there. Maybe you're thinking of alloys like which include gallium and indium?

        Only two elements are liquids at NTP, bromine and mercury. We get bromine from salt water mainly. And where do we get mercury from? H.G. Wells. --->

        1. Hero Protagonist

          Re: Another possible niggle

          “ And where do we get mercury from? H.G. Wells”

          Holy cats, I never realized H. G. Wells’ Mercury Theater was a pun on his initials until just now!

          1. Symon Silver badge

            Re: Another possible niggle

            Mercury Theatre was Orsen Welles, I think! But I see the connection...

  3. cb7 Bronze badge

    And another confusing point

    On the one hand the required pressure and temperature is "200 GPa at 1000K". But then it's superconducting at room temperature?

    What did I miss?

    1. Filippo

      Re: And another confusing point

      Nobody is claiming that it's superconducting at room temperature. The statement is that understanding metallic hydrogen better is interesting because it's a new bit of the conductivity puzzle, and solving that puzzle could eventually lead to room temperature superconductors.

      1. Chris G Silver badge

        Re: And another confusing point

        Actually, the last sentence in the article says,' Hydrogen also has a number of exotic properties including super conductivity at room temperature and superfluidity.

        I guess that prediction is based on being able to achieve sufficient pressure at room temperature.

        1. 0laf Silver badge

          Re: And another confusing point

          Boyles Law might have something to say about how comfortable it is in that room

        2. Anonymous Coward
          Anonymous Coward

          Re: And another confusing point

          I guess that prediction is based on being able to achieve sufficient pressure at room temperature

          That's my understanding too.

          The article refers to pressures of approximately 200 GPa at around 1000oC. At room temperature a proposed metallic hydrogen precursor has been created at a pressure of 384 GPa, so we have some idea of the even more mind-boggling pressures required to achieve room temperature metallicity.

          Nature magazine abstract:

    2. Jimmy2Cows Silver badge

      Re: And another confusing point

      Squash it enough*, say inside some kind of exotic high-strain metamaterial, and the theory is it should become superconducting a room temperature.

      Such a material is metastable at best, and likely to undergo highly explosive RUD with the slightest defect in the containment material. This has been shown with the diamond anvil experiments. Stable under the vast anvil pressure, but relax the pressure and boom.

      * For some vast, currently unachievable value of "enough". Think a lithium-ion battery or hand grenade is explosive? IThey'd be like firing a cap gun in comparison. The strain energy stored in the containment metamaterial would be many orders of magnitude greater, not to mention the compression energy stored in the metallic hydrogen.

  4. trevorde Silver badge

    Quantum computing

    We just need more cats

    1. Chris G Silver badge

      Re: Quantum computing

      "We just need more cats"

      Wanted, dead or alive!

      1. Jimmy2Cows Silver badge

        Re: Quantum computing

        Dead and alive. Simultaneously.

        1. Evil Scot

          Re: Quantum computing

          And Bloody furious!!!!!

        2. Alan Brown Silver badge

          Re: Quantum computing

          preferably carrying Tea and No Tea simultaneously

  5. Screwed

    Just yesterday, I happened to skim past an old QI in which it was asserted that all elements except H and He are classified as "metals" by astrophysicists.

    (Confirmed by a quick look at WIki.)

    I certainly remember chemistry/physics lessons in which it was claimed that H was somehow more of a metal than a non-metal. So this is not surprising.

    Does "metal" now have to mean "not He"? (Until someone demonstrates that too has metallicity.)

    1. Charlie Clark Silver badge

      Hydrogen is just odd and largely defies classification along with the other elements. It's an electron captured by a proton so it doesn't even have a proper nucleus except it has two isotopes that do. But it's metallic behaviour has been absorbed in many ways, not just because it's such an effective reduction agent but also because it can be adsorbed.

      Cool H and He enough and they get odder.

    2. Jimmy2Cows Silver badge

      Astrophysicists and regular physicists / chemists / us mere mortals apply different meanings to the term "metal".


      Regular physicists / chemists / mere mortals:

  6. PerlyKing Silver badge

    Neural net predictions

    I'm not well up on neural networks or machine learning, but from what I've read here and elsewhere, nobody really understands how they work. I can sort of understand how you can train a neural net to recognise various features in images, because you can test that. But I don't understand how you can use one to make predictions in this way - how do you verify the results?

    Can we have an icon for "thinking about this makes my head hurt"?

    1. Alister Silver badge

      Re: Neural net predictions

      But I don't understand how you can use one to make predictions in this way - how do you verify the results?

      I suppose it is analogous to the intuitive leaps that the human brain is capable of, and therefore the only way to verify the results is rugged testing of any prediction.

      This is very much why Artificial Intelligence is so much sought after - getting a computer to have the ability to extrapolate from too little data - but it does make me wonder whether humans will question and verify what a computer comes up with in the same way they would a human, or whether we will be inclined - as we do now - to just say "well the computer said so" and not question it.

    2. Paul Kinsler

      Re: Neural net - nobody really understands how they work.

      Actually, we know exactly how they work - it's specified by the architecture, the weights, the thresholding function, the back-propagation algorithm, the inputs, training data, etc (or, if you prefer, the code that implements it). The problem is, except for very very simple implementations, there is far too much information to understand, and the system is far too complicated to comprehend how - heuristically - we might get from some given input to a given output [1].

      It seems to me that it's more like an "Artificial Intutition" than an "Artificial Intelligence". It seems to work quite well in many cases, but there is no clear reasoning explaining why or why not, and so you never really know whether you can (or should) trust its answers (or, if you prefer, its proposals, guesses, etc) it or not [2]. But it's probably better than randomised guessing, ... except when it isn't.


      [1] Which is, of course, why we get computers to do it for us.

      [2] Thank goodness people are never like that. :-)

      1. Crypto Monad

        Re: Neural net - nobody really understands how they work.

        AI can also be easily fooled - such as when you change one carefully selected pixel in an image of a cat, and it classifies it as a dog.

    3. Jan 0 Silver badge

      Re: Neural net predictions

      It's based on the idea that neural network guesses might be better than human guesses.

  7. John Robson Silver badge

    That's not a simulation any more

    It's an "educated" guesstimate

    Where I use the work educated very loosely.

  8. Anonymous Coward
    Anonymous Coward

    Beware The Gosling effect

    Beware "Gosling effect" in AI:

    So, there is some force that binds the electron at some fixed distance. As the electron moves further away that force becomes a pull, as it moves too near it becomes a push. At the electron distance it is zero. Recognize you're dealing with a resonant force there, and that's the local resonant wavelength.

    Lets, for the sake of argument, commit heresy and assume Schrodinger's equation is an *approximation* for that system, and not the *definition* of that underlying system.

    If the randomness of Schrodinger is *complexity* instead of *uncertainty* then it's not that it cannot be predicted, its that you don't understand that oscillatory resonant motion, so you can only approximate it. But its complex not unknowable, you should be able to find repeatable cases, predictable cases. The existence of those repeatable cases is proof the system is not probabilistic, rather just that its complex and not understood.

    Even if you didn't understand the system itself, you might notice an electron returning to the same place in a predictable interval, or two seemingly unrelated things moving the same way, even if they're apparently unconnected, even if separated by time or space. That would tend to confirm that the motion is not a random-probabilistic motion, but rather a complex one you don't understand.

    You *do* keep finding these cases and you label them 'quantum teleportation' or 'quantum entanglement'.

    Labelling those things as "quantum teleportation", and claiming the properties are teleported from one to another (or to a future instance of the same thing) across another dimension borders on the ridiculous.

    You're dealing with an oscillating resonant system, and its complicated, yet somehow the electron and proton in the nucleus share the pattern, in order to be held at a resonant distance). That binding is so common and so fundamental, that the oscillation pattern for that binding must be pretty simple. So lets call that F2, where F1 is the underlying resonant pattern.

    1. Eclectic Man Silver badge

      Re: Beware The Gosling effect

      "That binding is so common and so fundamental, that the oscillation pattern for that binding must be pretty simple."

      Your reasoning, please. IN my experience, just because something is fundamental and common does not mean it must be simple.

    2. Glen 1 Silver badge

      Re: Beware The Gosling effect

      "borders on the ridiculous."

      No more ridiculous than requiring space to have curvature in "another dimension" to explain gravity. Yet relativity (and spacetime) is one of the most rigorously tested theories we have. Hell, GPS has to account for it to work.

      Quantum teleportation describes the *observed* phenomenon, much like dark energy and dark matter describe others. The theories as to *why* are... less clear

      Remember, the Schrodinger's Cat thought experiment was a *critique* of the superposition theory (Copenhagen Interpretation). After all, saying the cat is somehow dead *and* alive "borders on the ridiculous". Yet it matches the observed results. (the superposition, not the cat thing -YKWIM)

      *shrug* There are bigger brains than yours (or mine) working on it.

  9. JDX Gold badge

    >>Metallic hydrogen [also] has a number of exotic properties, such as superconductivity at room temperature and superfluidity. Understanding this material is thus useful for potentially utilizing this super materia

    The article makes it clear you cannot realistically produce it on Earth, which is rather the whole point of the clever software. So using it is rather a no-no.

    1. Cuddles Silver badge

      Not necessarily. At the moment, we don't really understand it at all, so the only thing we know about it is that it can be produced under the conditions at the centre of a gas giant. Part of the idea of studying things like this, aside from simply wanting to understand how things work, is to find out if there are other ways to make and/or use it that don't require such extreme conditions. You can look at something like superconductivity for a similar idea - it was originally discovered under very extreme conditions that were basically impossible to actually use in any sensible way, but from understanding things better we now make routinue use of superconductors in a variety of places. It took over 40 years to go from discovery to a practical application. Nuclear fusion would be another good example, although that one's been a bit less easy to get to practical use.

      So maybe metallic hydrogen will never be easily produced on Earth, and maybe it doesn't have any actual use even if we can. But on the other hand, maybe it will. We know we cannot realistically produce it on Earth right now, but without learning more about it, we really just don't know what might actually be possible.

  10. Cuddles Silver badge

    Since we can't build a planet

    To be fair, even with current technology we probably could. It would take a bit of investment of resources, so it would be accurate to say we probably shouldn't.

    1. Alan Brown Silver badge

      Re: Since we can't build a planet

      "To be fair, even with current technology we probably could"

      The problem is getting it to STAY that way.

      I would not like to be around the stuff when it decides to not be metallic hydrogen anymore

      1. Eclectic Man Silver badge

        Re: Since we can't build a planet

        Building a planet in our solar system would be very dangerous - it would have to be bigger than Pluto, Ceres, etc. to qualify as a planet, and therefore would have sufficient gravitation to perturb the orbits of nearby planets, such as the Earth, and asteroids, potentially endangering life on Earth by lumps of rock an ice falling on us. I vote against it.

    2. FrogsAndChips Silver badge

      Re: Since we can't build a planet

      Well, there's already this planet built by the Magratheans in the ZZ9 Plural Z Alpha sector, to conduct a very complex scientific experiment...

      1. Scott 26

        Re: Since we can't build a planet

        Took this long to have a HHGTTG reference..... for shame el Reg commentards, for shame

        1. FrogsAndChips Silver badge

          Re: Since we can't build a planet

          I was as surprised as you to be the first one to post it. Before reading the full article, I even thought the subheadline itself was a reference.

  11. Anonymous Coward
    Anonymous Coward

    Pluto... not a planet (grrrr)

    Well then hydrogen ought not to be considered a significantly sized element.

    1. Eclectic Man Silver badge


      Don't 'grrrr' at me, it was the IAU wot decided to demote it to minor planet.

  12. Conundrum1885 Bronze badge

    Re. Metallic

    Actually it was Ashcroft who wrote the equations predicting that hydrogen might metallize and have strange properties.

    That he didn't receive a Nobel yet is most unfortunate as its well deserved.

    (here's hoping this year)

    1. Conundrum1885 Bronze badge

      Re: Re. Metallic

      Apologies, forgot to link the article.

      (fx: Liquid Metal facepalm)

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2020