I am trying to understand how useful liquid metal is, this means what happens when liquid metal is applied in different scenarios. I find a lot of conflicting/contradicting statements, also in this thread of 8 comments and would like to explain my current model of understanding by highlighting where I agree, but also to explain where my view is completely different together with the reason or reasons I see it completely different.
Ideal would be to get some feedback from people with a lot of knowledge and practical experience or advice to another thread where liquid metal is or was discussed quite deeply to improve my model of understanding.
I fully agree to the first quote. Due to its quite high thermal conductivity of 40 to 80 [W/m-K] in combination with possibly very thin layers (when both surface forms fit well and there are no mechanical limitations like the U-form of a IHS/lid) only solid metals as silver/copper/aluminum/… can transfer/conduct heat better than liquid metal (but solid is not really usable for gaps that might change their form with temperature or are showing different thermal expansion), while materials like graphite or micro-diamonds with even higher thermal conductivity than solid metals each have their specific problem (graphite is not yet {or ever?} possible to realize in the thinness of thermal grease/liquid metal and micro-diamonds cannot realize a large contact area and their hardness causes surface damage when pressed onto/into solid surfaces).
Concerning the second quote I agree to its first part, due to immediate, but only thin oxidation the surface exposed to air gets crunchy. But I think think this oxidation of gallium on air has no real effect on the thermal resistance {it would only be harmful if oxidization would continue}, but surface oxidization influences strongly how well a surface can be wetted (how low the contact resistance of 2 different materials is) as it results/support to form the ball forming of liquid metal.
But in my try to understand what is happening around liquid metal I do not agree to the second part of the first quote. In case of non-conductive thermal grease I understand the dry out has something to do with the de-composition of the mixed materials and evaporation of the material into air that keeps the grease flexible (this means in case of thermal grease sealing to reduce the evaporation into air would have an effect on longevity).
In case of liquid metal however evaporation into air would start only at temperatures above 1.000°C, but "dry out" hardening is rather the result of migration/diffusion of gallium into copper to form the alloy CuGa2, which you find on copper after a longer exposition to liquid metal as a mechanically stable thin silver colored surface. This thin CuGa2 layer results in negligible worse thermal temperatures, because even though the thermal conductivity is reduced to 95 [W/m-K] for the alloy CuGa2 compared to about 400 [W/m-K] for pure copper, the thickness of the CuGa2 layer is extremely thin. The "dry out" hardening of liquid metal is in my understanding caused by the fact, that reduction of the gallium percentage in the liquid metal (because the gallium disappears/migrates/diffuses into copper and at much slower speed also into nickel) will increase the melting temperature, so that first a little of the liquid metal sets down as solid metal on the surfaces and with steadily migrating/diffusing away gallium eventually all remains of the liquid metal will become hard at room temperature causing large temperature up on the silicon chip either by the creation of a large void in the gap (steadily less contact area) as the liquid metal volume becomes less and less or at gaps having excess liquid metal it gets first completely hard and then cracks under the mechanical stress of repeated thermal changes. Sealing cannot have any effect on the "dry out" hardening process because it happens based on migration/diffusion of a liquid metal/gallium into a solid metal it cannot be sealed off. The main factors for the speed with which liquid metal "dries out"/becomes hard should be: 1) the material the liquid metal/gallium is in contact with (migration/diffusion into copper is fast, into nickel is slow and into silicon is barely noticeable if happening at all) 2) the thickness of the liquid metal gap (the thicker the longer it takes to dry out), 3) the size of the contact area (porosity or fine scratches result in larger effective contact area, faster dry out) 4) the temperature (the higher the temperature, the faster happens the diffusion).
When my basic understanding on the completely different dry out processes on liquid metal compared to thermal grease is correct, this should also mean that a high pressure heat sink (resulting in lowest thickness liquid metal) should dry out faster, simply because the speed of gallium to leave/form CuGa2 will be the same, but the amount of gallium is much less.
When you put on the other side some storage in form of a foam ring around the liquid metal the time to dry out should take longer, because the gallium migrated into the copper will partly be replaced from gallium/liquid metal stored in the foam, as long as both stay in contact.
Out of above model of understanding my personal conclusions on liquid metal are:
1) Liquid metal is well usable on nickel (Ni) plated copper (require no or rare reapplication/renewal of the liquid metal due to "dry-out" hardening, because gallium's speed to migrate/diffuse into nickel is low).
2) Liquid metal is usable/effective on copper, but because the speed of "dry out" is much faster than on nickel, you have to accept to replace the liquid metal each half a year (or whichever time your specific application process/surrounding allows the liquid metal to stay liquid/be removed effortless).
In addition:
3) Liquid metal has the general problem of leakage (this means danger of catastrophic PC/graphic card failure by causing either electrical short or in longer term destroying solder joints), because it is of low viscosity and nobody can control the amount applied, the contact pressure applied and the changes of forms by temperature in such a way that no spillage will occur. This means precautions to prevent short/contact are absolutely mandatory and good security is given by following 2 step precaution:
a) Coating of close by capacitors/resistors by either nail polish or any other coating (should be long term stable against liquid metal) or a tape suitable for at least 100°C (Kapton, …).
b) Setting a barrier around the liquid metal so that spilled small balls cannot roll to/reach any dangerous point (for example a compressible foam ring {on a CPU the lid acts as such barrier}).
4) Liquid metal shows great reduction of temperatures (-10/-20°C) at high load (this means chip temperatures close 90°C) only when it is applied as first TIM directly onto the small surface silicon chip, while when applied as second TIM on top of an IHS/lid to an air or water based cooler its effect is much smaller (-1/-3°C). Seeing the need for permanent renewal of the liquid metal and its required precautions only extreme enthusiasts looking for highest overclocking may apply liquid metal on top of an IHS/lid while most people will use for the second TIM the much more easy and simple to use thermal grease accepting the 1 or 3°C higher temperatures to get much longer longevity {not required to reapply each 6 months and no trouble with leakage causing a short or destroyed solder joint}).
The oxidation of liquid metal on air is in my understanding the reason why it is absolutely necessary to wet both opposite surfaces. When both surfaces are wetted the oxide cannot harm, while when only one side is wetted, the thin oxide layer on the liquid metal may hinder most area on the opposite side to get wetted which results in strongly higher thermal resistance.
Some comments on detailed wordings in the first 8 comments in this thread, which I think are misleading for people with few knowledge or even show completely wrong understanding:
Instead of "He could have re-pasted every 6 months or a year to keep it "fresh"" more correct wording should be "Liquid metal on copper MUST be replaced every 6 months or whichever time a specific attachment process keeps the metal liquid/easily removable".
Instead of "Don't screw with Liquid Metal, just get a nice non-conducting paste and be happy with a few degrees less cooling" the sentence "If you do not want to replace liquid metal on top of an IHS/lid each 6 months to get just 1 or 3°C down, accept the slightly higher temperature and go with thermal grease offering much longer longevity than liquid metal and no danger of catastrophic short damage" would be much clearer.
The sentence "Simply drying out though is usually due to insufficient pressure" is in my understanding nonsense, as higher pressure has basically no effect or rather the contrary effect (as it reduces the available gallium amount to the minimum, a shorter time to dry out is the consequence).
When someone is determined to use liquid metal as "I have always found copper heatsinks need to be sanded and redone about twice per year" I personally would question the process to sand the copper heatsink. Clearly the remains of the liquid metal must be removed, but this should be done without sanding (rather heat up to liquify the hard remains or use a soft acid/stop) to leave the silver color CuGa2 layer on the copper as this means this thin CuGa2 layer may reduce the speed gallium can migrate from newly applied liquid metal more deep into the copper (behind the CuGa2 layer).
On the out of theme sentence "Heck we don't even have active cooling for NVME controllers for Gen3 or Gen4 Storage. There's a lot of engineering needed to clean up laptop thermals." I think current Gen 3 controllers are at the limit and PCIe Gen4 storage will only be widely applied once the efficiency of the controllers has increased to a level that allows usage without active cooling, clearly the preferred way for laptops.
With my model I can explain most statements made above or elsewhere. But for a sentence like "
Papusan got something like 3 or 4 years (don't remember exactly how long) out of one application of liquid metal on his Alienware 17 (Ranger)" my model does not provide an easy explanation. A longevity of 3 or 4 years on copper is a much longer time than I would expect seeing a general saying to replace liquid metal after 6 months (and some people say they even experiencing high temperatures in less than 6 months). But trying to explain this unusual long time by "better fit" is not really convincing me as my model says "better fit" means less gallium available, this means shorter longevity. Many people inform too few facts because they look for only one main cause of influence. In case of liquid metal however several causes can add up and result in a quite strong deviation from a general recommendation, like quite low average temperatures/few use, rather thick liquid metal when applied on a silicon die and a low reaction speed surface condition could result in such extreme deviations when all added up in an the same direction.
My point is, most people see clearly liquid metal is not allowed on aluminum because the alloy gallium forms with aluminum is mechanically very brittle/has extremely low thermal conductivity and gets eaten away by air, but they do not see that gallium also migrates/diffuses into copper (and at less speed into nickel) which in turn reduces the gallium percentage in the gap resulting in "dry out" hardening as the melting temperature of the remaining liquid metal gets higher and higher with the ever lower percentage contents of gallium.
Last edited:
Mar 28, 2020
Problems? See this thread at archive.org.
