What would happen if you put a vast amount of liquid metal into a custom cooling loop instead of water/coolant? What challenges would you face? Would there even be any benefit to doing this?

BONUS: What if you used copper tubes instead of standard plastic/glass tubes and pumped liquid metal through the copper tubes? And also used a copper CPU block as well?

  • Comments are not for extended discussion; this conversation has been moved to chat. – Journeyman Geek Mar 2 '18 at 6:09
  • How many metals are liquid at room temperature? All of the other metals would need to be heated to a high temperature, which would heat up—rather than cool down—your system. – Infiltrator Mar 2 '18 at 6:24

Everything in Keltari's answer is right, I just want to expand it with some other important info:

When you want to "transfer" heat, you need to deal with 2 major values: Thermal conductivity and heat capacity. First is how easily get/give heat from/to other material, like get the heat from hot surface and give the heat to cold surface. The second is how much energy can it store.

Thermal conductivity of liquid metals are very low compared to solid ones. Pure, solid, aluminium has a thermal conductivity of about 200 W/(m K), pure copper is about 390 W/(m K). Mercury, on the other hand, has a value about 8.5 W/(m K) and the value for water is about 0.6 W/(m K). So liquid metals are better than water for heat transfer, but much worse than solid metals.

The heat capacity is another part. A 1 K change in temperature (i.e. 1 °C or 2 °F change) for liquid water requires 4.187 kJ/kg, while the same change for mercury is 0.125 kJ/kg, this means same heat from the CPU surface incurs a 32 times larger temperature change in mercury!

If we think simply, 14 times better conductivity and 32 times worse heat capacity is about 50% worse sum related to water cooling, and still not taking into account other dangerous factors, like toxicity or the short circuit factors. (This calculation is not proper, because there's many other parameters which these values depend on, such as current temperature, pressure, and there is side dissipation on transfer, etc.)

  • What if theoretically you had some sort of copper heat pipe tubes that pumped liquid metal instead of water? See revised question. – FatalSleep Mar 1 '18 at 3:08
  • 3
    @uDev Thermal conductivity of NaK is 218 W/m, slightly better than aluminium. The problem is not with metal being liquid, it's with choosing mercury which is the worst metal for cooling. It's like saying that solid metal is unsuitable for radiator because titanium is only 21.9 W/(m·K). This answer is based on false premise. – Agent_L Mar 1 '18 at 19:01
  • 1
    I'm far from an expert on the subject, but it seems to me that the lower heat capacity of mercury could be overcome increasing the flow rate. – canadianer Mar 1 '18 at 19:55
  • 1
    @uDev Presumably the cooling system is transferring the heat to the environment, in which case the heat capacity of the coolant itself shouldn’t matter so long as the flow rate is high enough. – canadianer Mar 1 '18 at 20:27
  • 1
    @uDev The Question asked what would happen if you replaced water with a liquid metal. Your answer is partially incorrect and needs to be edited to reflect a volumetric comparison of heat capacity. Mercury has ~14 times the conductivity and ~44% the heat capacity of water. I think this means that maximum amount of heat that a Mercury system could move is ~7 times the maximum amount that could be moved with water given the same amount of time. – Kenneth Moore Mar 1 '18 at 22:51

While on the surface this might seem to be a good idea, in actuality, this is a very bad idea.

There are two metals (not including alloys) that are liquid at room temperature: Mercury and gallium.

First off, mercury is extremely toxic and should only be handled by experts.

Gallium will corrode aluminum and steel, which is what the coolant runs over/through to sink heat. It will eventually destroy the joints and heat sinks, which will lead to the next problem.

Both mercury and gallium are electrical conductors. If either of the two liquids were to leak onto the electronics, it could cause shorts and even damage the electronics. And again, mercury is extremely toxic. This alone is a reason not to use them.

Mercury and gallium have a high rate of volumetric expansion due to heat. Under high heat, they can expand greatly and the pressure would destroy the cooling lines.

Gallium itself isnt a liquid at room temperature. It has a melting point of 85.58°F (29.76°C), which means of the PC was turned off and it completely cooled, gallium would solidify. This of course could cause problems, since the liquid would not be able to flow.

Editing in some more thoughts:

Mercury is very, very heavy. One liter of mercury weighs a hair under 30 pounds (13.5 kilograms). One liter of gallium weighs 13.02 pounds (6 kilograms). It would take a massive pump to move that liquid around. The weight alone could cause PCBs to flex or break.

  • 4
    I thought organic mercury compounds were extremely toxic, and mercury itself is only moderately toxic? (i.e. you eat it and die, but touching half a drop of it with your skin won't kill you - unlike the aforementioned organic mercury compounds) – user253751 Mar 1 '18 at 4:05
  • 3
    "Gallium is corrosive to all metals except tungsten and tantalum, which have a high resistance to corrosion." Does Gallium (liquid or solid) corrode all forms of brass? – DavidPostill Mar 1 '18 at 6:11
  • 3
    volumetric expansion can be dealt with using a reservoir that isn't filled to the brim (iow give it somewhere to expand to) – ratchet freak Mar 1 '18 at 11:02
  • 5
    Ricin and tetrodotoxin are extremely toxic. Mercury is "handle with care, do not eat." – hobbs Mar 1 '18 at 17:03
  • 5
    Elemental mercury is not even remotely close to "extremely toxic". You can handle it and throw in a trashcan. You can eat it and it'll come through the other end. You can inject it, and it won't hurt you in ways other than mechanical blockage. You'll be fine, as long as you don't do it repeatedly, because bioavailability of elemental mercury is very low. Dimethylmercury, on the other hand, you touch a drop of it with a rubber glove and that was your lethal exposure. – Agent_L Mar 1 '18 at 17:35

Liquid metal CPU coolers already exist:


This one uses NaK : a eutectic alloy of sodium and potassium, that is frighteningly reactive with air, water, and just about anything:


The same alloy is used for cooling in the nuclear power industry.

  • Anyone who took high school chem would know that's a very, very bad idea for PC cooling utilization... all alkali metals are highly reactive, and KNa isn't any different. Combine that with the EM radiation, and iron casing to contain it, really makes such a product impractical for consumers and work stations alike. Cool invention, serves a purpose, but I can't imagine a home or business PC being outfitted with one of these, especially considering the added insurance cost for renter's/home/business insurance due to the high risk hazard of alkaline metal reactivity. – JW0914 Mar 1 '18 at 19:46
  • 1
    @JW0914 I think you may be vastly overestimating the risk. Consider that people are allowed to have ovens and stoves in their house. They're even allowed to have gas stoves, which are hooked up to pipes that can deliver unlimited highly explosive fuel. And don't get me started about those death traps they keep in their garage that have 15+ gallons of gasoline stored in them! – Cort Ammon Mar 2 '18 at 1:33
  • @CortAmmon Fuel requires an ignition source, alkali metals do not... the exothermic reaction that occurs with alkali metals is enough to spontaneously combust the metals in a violent explosion (if you never experienced this in high school chem, check out YouTube). This is a cool product, but carries a level of risk most informed consumers (let alone insurance companies) would not find acceptable. – JW0914 Mar 6 '18 at 2:16
  • @JW0914 I've seen what a few grams of alkali metals can do. It's impressive, don't get me wrong. But it's impressive on quite a small scale. I'll put it this way, I think insurance companies should be far more concerned with toddlers than they are with Danamics heat sinks. Toddlers don't need an ignition source either. – Cort Ammon Mar 6 '18 at 2:47
  • Or, for that matter, owners of Samsung tablets. It's amazing the hazardous things we carry around on our person! – Cort Ammon Mar 6 '18 at 2:50

Would there even be any benefit to doing this?

No. WC loop is not your central heating loop which works on temperature gradient. In a typical, properly sized WC loop, the coolant is circulated fast enough that all elements (blocks and radiator) are at almost same temperature. This means that better coolant wouldn't change much, and the entire loop is limited by the radiator performance. Even if so, as Nat said, heat transfer by coolant is [heat capacity] * [flow rate]. So it's difficult to overstate how much easier is to replace the pump with something from Laing E series (and change tubing to larger to keep friction low) rather than design everything up from scratch for a liquid metal coolant.

Even in nuclear industry, liquid metal is used not just because it has more heat capacity than water, but because water has neutron-moderating properties which makes it totally no-go for fast neutron reactors (like the one onboard USS Seawolf).

BONUS: What if you used copper tubes instead of standard plastic/glass tubes and pumped liquid metal through the copper tubes?

Nothing. The speed of heat transfer along a copper pipe is insignificant compared to the speed of heat transfer via the moving coolant inside. Just as with heatpipes. They're copper to move the heat in and out. Longitudinally, heat is moved by vapor - that's why once punctured, heatpipe becomes useless.

And also used a copper CPU block as well?

Most of them are copper already. If that's not obvious, it's because they're nickel-plated.

If you want drastic improvement in WC performance, move the radiator to a cold place, like out of the window. 16°C stress is easily doable in winter : ) Keeping radiator in same airflow as other components nullifies the largest advantage of WC: moving the heat far, far away.


This sort of thing could be fairly hazard-prone and seem to be a major safety issue for someone trying it at home. So, seriously, this response is hypothetical - don't try any of this at home, etc..

@uDev's answer is correct that you'd be primarily concerned with two things:

  1. thermal conductivity: How fast thermal energy (heat) moves through the substance.

  2. heat capacity: How much thermal energy (heat) a substance can hold (in this case, before it's too hot to absorb anymore).

Water's often a great coolant because it has a pretty high heat capacity. This is, it takes a relatively large amount of heat to warm it up.

That said, I think that some of the other answers overestimated how important heat capacity is in this case. The issue's that we're not really just heating up a set amount of coolant; instead, the coolant's constantly flowing, such that we're basically concerned with

  • [heat capacity] * [flow rate].

So if a coolant with a lower heat capacity is selected, the difference can be compensated for by increasing the coolant flow rate, up to some reasonable limit, e.g. where the frictional heat of the fluid flow becomes problematic or the pressure of the flow causes mechanical damage.

So, yes, in principle the greater thermal conductivity of a liquid metal might be helpful in some designs.

A practical limitation is that the cooling loop provides only one source of thermal resistance in the cooling mechanism. So, even if it were optimized to have a very low effective thermal resistance, the overall system's thermal resistance could continue to be propped up by the thermal resistance of the CPU and the heat exchanger on it.

  • There's lot of parameters what I not talked about. Thermal engineering is a separated science, cannot fit in some lines. When I planned a sigle laser diode cooler (about 10 years ago, from diode integrated case until cooler+fans), it took weeks (inclusing confirmation experiments) until we get acceptable solution (which is still in production). – uDev Mar 1 '18 at 10:54
  • 1
    @uDev Hah yeah, these systems can have a lot going on. Honestly the question seems like a bit of a poor fit for SuperUser since this site apparently doesn't have TeX enabled; it'd have been more fun to address it on SE.Engineering or something. – Nat Mar 1 '18 at 11:21

Not the answer you're looking for? Browse other questions tagged or ask your own question.