Liquid Metal Explained: How it works, why it fails (and how to use it)

Alternatively titled: 'Beginner's guide to liquid metals' OR 'Do liquid metals really damage copper? Yes, they do'

There's a lot of uncertainty out there about using liquid metal (LM) as a thermal paste: what is it, how do you use it, and can it cause lasting damage or corrosion? The temperature benefits of liquid metals are undeniable, but chances are you've had questions about their durability and safety over the long term.

Needless to say, some basic googling digs up countless differing opinions (1)(2)(3) about this topic, which means getting clear answers can be difficult.

The point of this thread is not to discuss the effectiveness of liquid metals (very effective), nor whether they pose any risks (they do), but rather to explain why the applications of liquid metals yield the results that they do.

Hopefully, this information is insightful and helps you decide whether liquid metals are right for you.™


What is Liquid Metal made out of?

Spoiler: Click me to hide/unhide

The principal component of liquid metal is Gallium, a soft metal that melts at slightly higher than room temperature (29C). It's nontoxic, unlike mercury. When combined with Indium and other metals, the melting/freezing point of the finished gallium alloy drops to nearly -19C. This means that at normal temperatures, liquid metal remains liquid. (And it shouldn't evaporate significantly, since the boiling point is 1300C[1])

To be more accurate, the liquid alloy is called "galinstan" - and the exact ratio of gallium, indium, tin, and other metals is proprietary.

When you buy liquid metal, whether it's coollaboratory liquid ultra (CLU), thermal grizzly conductonaut, or just straight up galinstan - you don't know exactly what's inside. (And there is reason to believe that the formulations are different since CLU and conductonaut have different viscosities)

But regardless, any liquid metal brand works well as a thermal interface material/thermal compound (TIM) because the stuff is temperature stable and has a high thermal conductivity of 16.5 W/(mK)[1] {versus solder at 32-94 W/(mK), and corning TIM at 0.5-3 W/(mK)}. [2]

Unfortunately, the problem with LM is that it's electrically conductive. Combined with its very unique consistency, this makes LM a potentially difficult material to work with. And - as we will see - LM is very reactive to aluminum, and erodes copper and nickel to a lesser degree. Be careful with both how and where you apply it.

Comparing Liquid Metal versus Thermal Paste

Spoiler: Click me to hide/unhide

Conventional thermal paste

• Usually not electrically conductive (don't need to worry about causing short circuits)
• Worse at transferring heat
  • Claimed 0.5-6.0 W/(mK) thermal conductivity (more is generally better)[2]
• May degrade over time
  • Example [link] of dow-corning's TC-5022 performance over 20,000 on-off cycles. Other pastes vary.
• Doesn't chemically damage heatsink surfaces
• Easy to apply

Liquid metal

• Electrically conductive (will cause short circuits)
• Better at transferring heat
  • Claimed 16.5 W/(mK) thermal conductivity (more is generally better)
• Damages heatsink surfaces
  • Causes degraded performance over time
• More difficult to apply
  • Must be brushed onto surfaces (or risk drops running off and causing shorts)
    
  • Must apply thin layer (or risk drops running off and causing shorts)
  • Must insulate contact area (or risk drops running off and causing shorts)

You are already probably familiar with the consistency of thermal paste, it can vary from gooey (most thermal pastes) to dough-like and hard to spread (IC diamond).

What's more interesting is the consistency of liquid metal: it is a liquid with high surface tension that can also be 'brushed' into a thin film.
Unlike conventional thermal pastes, liquid metal must always be spread manually onto the contact surface.
Also unlike conventional thermal pastes, liquid metal is not compatible with aluminum. Never use liquid metal on surfaces that will contact aluminum.

(Yes, it really squirts out of the tube like this)

Where can Liquid Metal be used?

Spoiler: Click me to hide/unhide

Liquid metals can be used on desktop CPUs (between the die and IHS, or between the IHS and heatsink), on desktop GPUs (between the die and heatsink), or in laptop CPUs (between the die and heatsink) and laptop GPUs (between the die and heatsink). Note: the heatsink in these cases must always be copper, never let liquid metal contact aluminum.


First, some background information:
Liquid metal has most often been used for "delidding" intel desktop CPUs.

To understand what delidding means, take a look at this not-to-scale diagram of a typical desktop CPU that's been sliced in half. (We'll compare this to laptops in a second)

Heat from the CPU has to travel from the silicon die (grey in diagram above) through a thermal interface material (TIM), through a heatspreader (IHS), through another TIM, and finally to the heatsink. That's a lot of interfacing so we want to minimize thermal resistance as much as possible.

Unfortunately, intel has stopped using solder (a metal alloy) between the die and heatspreader (IHS) since the Haswell generation, instead preferring thermal garbage™ (to save on manufacturing costs. There was some noise about how toothpaste is safer since solder can cause die cracks due to differing thermal expansion, but that is nonsense considering how AMD and intel both have used solder for decades).

The process of delidding refers to physically removing the IHS, scraping out the toothpaste, and reapplying a better material such as liquid metal. Then the IHS goes back on like normal. Result? Way better temperatures and more thermal overclocking headroom.

When desktop CPUs are delidded, liquid metal is most often applied between the silicon and the nickel-plated copper heatspreader.


On laptops there is usually no reason to delid a CPU since mobile chips do not have heatspreaders. (There are exceptions: some laptops use socketed desktop CPUs)


The heatsink directly contacts the die through a thermal interface material. Often, laptop heatsinks are finished with worse quality than desktop CPU heatsinks and since there is no heatspreader the thermal interface material serves as the sole interface between the hot die and the heatsink. Therefore, it's super important that we use good thermal compound and apply it well.

Unfortunately, good thermal paste is not always used, and often it is applied poorly or with great variance. By using a better thermal paste - or liquid metal - and reseating the heatsink, we can achieve noticeably improved temperatures.

Repasting with liquid metal on a laptop means applying it directly between the silicon and the copper heatsink.

Why consider Liquid Metal?

Spoiler: Click me to hide/unhide

Generally, for the same reasons you would consider reapplying the thermal paste to begin with.

If your thermals are out of whack, CPU/GPUs are throttling, or your fan speeds are way higher than normal - it might be time to consider a repaste.

The improvements from liquid metal are pretty substantial, I wont even link to results because you can find them yourself. Obviously, while you can get good results from a conventional thermal paste, liquid metal tends to do even better (if you are looking for a conventional paste, thermal grizzly kryonaut is one of the best and is very easy to work with).
This is because of simple physics: assuming all other variables are constant, if your thermal material has a higher thermal conductivity then your thermal interface will have less resistance.[2] As a result, you transfer more thermal energy out of the system (CPU) and into the heatsink per unit of time. This drops your CPU temps and increases your heatsink temperature, and the increased delta T of your heatsink vs ambient air means more instantaneous cooling. (see: newton's law of cooling)

If your thermals just seem okay and your laptop is hard to take apart (check your service manual), repasting may not be worth the trouble.

Some food for thought:
Many people don't consider the fact that same-model laptops may be cooled differently because of manufacturing variation:

1. Variation in heatsinks (roughness of the contact plate, are the heatpipes bent?, is the contact plate bent?)
   
2. Variation in thermal paste application (amount of thermal paste, position of thermal paste, mounting pressure, thermal pad placement, quality of paste used, etc)
3. Variation in silicon (differences in operating voltage, silicon characteristics [leakage, ASIC quality])
4. Variation in components (choice of board components: VRAM? mosfets?)

The result is that while you and the guy next to you may have the same model laptop from the same factory line, one of you might get noticeably better thermals under the same loads.

Then there's the fact that laptops tend to get hotter (and thus louder) over time, due to

1. Dust buildup in vents & fans (reduces airflow)
2. Seepage & wearout of thermal paste[2]
   
3. More installed software (correlated with processor load, reduced C-state residency = more power & heat)

Some tricks to consider before reapplying thermal paste/liquid metal:

1. If you have an intel cpu, haswell or newer: try throttlestop to undervolt your CPU which reduces power use.
   
2. Discrete GPU users: Try MSI afterburner to undervolt your GPU (not always possible)
3. Try blowing the vents clean of dust

Dangers of Liquid Metal:
Electrical Conductivity

Spoiler: Click me to hide/unhide

Before you try working with liquid metal for the first time, please always remember that liquid metal is metal, it will conduct electricity.

If you spill this stuff on your laptop's motherboard (or squeeze some of it off the CPU/GPU when you screw down the heatsink) and then turn the computer on, it could short circuitry.
RIP computer. It happens.

This is why you must use small amounts of LM and seal off the surrounding area.
Example: http://forum.notebookreview.com/threads/questions-about-liquid-metal.803973/page-3#post-10535148
Sometimes there are surface-mounted components on the CPU/GPU PCB. Some people use electrical tape to cover those components, others a coat of varnish, others kapton tape, etc. In any case, your seal around the liquid metal should be as close to airtight as possible.

Dangers of Liquid Metal:
Reactivity with Copper Heatsinks

Spoiler: Click me to hide/unhide

It's been debated whether liquid metals damage copper heatsinks.
They do: over time, the gallium in liquid metal will be absorbed into the copper heatsink, causing the LM to "dry" out.

Let's explain in more detail:
The electrode potential of gallium is -0.53V, nickel is -0.24V, and copper is +0.34V.
The difference between gallium and copper favors a reaction that occurs even at room temperatures.

Obviously, all liquid metals have a high gallium content (plus other metals to reduce the melting point). When the gallium contacts pure copper, the metals irreversibly alloy. This reaction proceeds until there is no more copper or all the gallium is consumed [3].

The reaction is: Ga + Cu → CuGa2 [67%] + Cu3Ga [11%]. ( + Ga2O3 [12%])
Both CuGa products are stable until 175C[3][4].



The means liquid metal will literally eat into the copper until the gallium is gone, and the resulting copper-gallium alloy is a silver-ish color. Yes, - in case you are wondering - the gallium in liquid metal reacts this way despite the fact that there are other metal stabilizers present in LM[5].

The non-gallium components (indium, tin, etc) of the liquid metal[3] which are solid at room temperature will be left behind on the heatsink surface as this process occurs. The formation of this non-gallium metal deposit is most obvious visually when the gallium is totally absorbed into copper. Do note that this residual non-gallium liquid metal is hard and brittle, as you would expect. While this deposit is technically metal and is a good heat conductor, it does not form evenly and therefore it's highly likely that an air gap will also form between the die and the heatsink, and your laptop will hit thermal runaway at this point. This video [link] is a good example of the consequences of this process. This mechanism appears to be the most common cause of long-term failure in LM applications.



Note that at higher temperatures, the invasion of liquid gallium into the copper heatsink only gets faster.[6] Anecdotally, it appears that this process can take anywhere from just a couple months to a year+ until a point of failure is reached.

The factors influencing the speed of this process include obvious ones like temperature, formula of LM used, surface roughness, and amount of LM used. But, the porosity and purity of the copper heatsink may also play a role. Due to all these variables, accurately predicting the rate of erosion for an application of LM is simply not possible.
Here's a graph of the mass fractions of the obtained CuGa alloys at various temperatures (oxygen atmosphere) for you nerds: [link]

This effect is less observed in the classic delidded desktop CPU because the gallium in the liquid metal is far less reactive against the nickle plating of the CPU heatspreader. (The nickle plating is designed to protect the copper against normal solder alloying, but also happens to be effective vs gallium[7]).
If some people tell you that LM 'dries out' while others say it's totally stable, now you know why. LM is fine under a CPU IHS, and its even fine when used between a die and pure copper, but in long-term use it will pit copper surfaces and this can lead to temperatures that stay stable for months but suddenly spike towards the end of the LM's usable life. Again, nickel surfaces are also pitted, but just at a significantly slower rate:

This image: coolaboratory liquid pro after 1 year on a nickel-plated copper heatspreader. The excess LM has been removed to reveal the heatspreader surface. Gallium-Nickel alloying is clearly visible, but the thermal impact of this alloying is likely minimal. If this surface were exposed copper instead of nickel, then the damage would be worse and you may be able to see metal deposits on the surface.

The last interesting things to note are:
1. At 20C, the thermal conductivity of CuGa2 (the principal alloy of gallium and copper) is 98 W/(m⋅K)[1], while copper's thermal conductivity is 400 W/(m⋅K).
2. While we don't know the exact formula for any of the liquid metals, they are all gallium based so they will all attack copper to varying extents.

Conclusions:
1. Liquid metals will visibly degrade the copper heatsink surface over time.
2. Simply buffing the residue off the copper heatsink and reapplying the LM might actually be OK. The CuGa alloy obviously can't match pure copper for heat conduction but it's still way better than solder or thermal paste - the formation of copper-gallium alloy alone should not be the cause of thermal bottlenecks.

It's not clear from my research how deep the gallium attacks into copper. It is clear that LM will alloy with copper and 'dry out'. However, if over multiple applications the gallium can't penetrate its own pitting anymore, then the LM invasion into copper will stop and you can - theoretically - end up with a 'stable' LM application.

If your laptop's heatsinks have copper contacts to the die and the inconvenience of potentially opening up your laptop to reapply liquid metal every 6-12+ months is okay to you, then liquid metal should be fine.

Dangers of Liquid Metal:
Reactivity with Aluminum

Spoiler: Click me to hide/unhide

TODO

How to use Liquid Metal

Spoiler: Click me to hide/unhide

TODO

Frequently Asked Questions

Spoiler: Click me to hide/unhide

TODO

Sources
1. https://sci-hub.ac/http://ieeexplore.ieee.org/abstract/document/6231443/?reload=true
2. https://i.imgur.com/QAOaxtg.png (slide from dow corning TIM presentation)
3. http://www.ipme.ru/e-journals/RAMS/no_81808/grigoryeva.pdf
4. https://sci-hub.ac/https://www.scientific.net/DDF.326-328.227
5. https://www.osti.gov/scitech/servlets/purl/811932
6. http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1124&context=mechengdiss
7. https://overclocking.guide/the-truth-about-cpu-soldering/

Please feel free to comment if there are any mistakes, my reading over the papers linked above was pretty cursory so I welcome anyone who can help clarify. This stuff is all based on questions I had personally (mainly "is this safe to use on a laptop") so I hope this can be useful for others in some way.

As always, feel free to leave feedback or ask questions, that's what the forum is for.

Thread is a work in progress. 9/27/17

More reading material for bored people:
http://forum.notebookreview.com/thr...y-liquid-ultra-any-tips-before-i-start.741745

 

Good read, I like it.

 

Nice work, thanks, it's really informative.

 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5772911/

 

Definitely a great guide for the beginners. I would like to add something in regard with using Liquid Metal thermal paste on GPU. As far as I have experienced, i never recommend using liquid metal with GPU, there are insane chances of shorting everything and end up with damaged GPU. Even the good thermal pastes like the Grizzly Conductonaut are not safe to use untill you know what you're doing and the right process.

 

Grizzly Conductonaut = Liquid metal

 

ha wow i sure didnt expect an NCBI reference here

good job OP, keep it up

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

SOrry if I am necroing this, but I bought a clevo with a delidded cpu and during the purchase selected CLU to be applied between the heatsink and the HIS.

Long story short, do I need to clean the LM everytime I remove the heatsink like regular paste or is simply putting it back in place enough?

 

simply putting it back would result in very horrible temps. as with any kind of TIM, it needs to be cleaned off and freshly reapplied every time the heatsink is taken off.

also, are you absolutely sure the LM was applied between heatsink and IHS, not cpu die and IHS? afaik, vendors only do LM beneath the ihs, not on top of it due to too much risk involved for it to spill out during transport.

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

My laptop is from Eurocom and they confirmed that between the die and HIS they were using conductonaut and I had the choice between IC7, Kryonaut and CLU and I selected CLU for the heatsink and HIS.

Thanks for confirming, guess that if I want to repad and repad for the gpu side I'll have to wait.