Something to think about, liquid metal compatibility with copper heat sinks

Discussion in 'Hardware Components and Aftermarket Upgrades' started by Tishers, Jan 27, 2017.

  1. Tishers

    Tishers Notebook Consultant

    Reputations:
    90
    Messages:
    105
    Likes Received:
    196
    Trophy Points:
    56
    While those who are familiar with galinstan (liquid metal) type thermal compounds and how very bad it is to apply these to aluminum it is probably not thought of of how this works with copper.

    Galinstan is an alloy of gallium, indium and tin that has a melting point of -19 C and a boiling point of 1300 C (no other metal has such a big difference between the melting and boiling point). This means that it remains a liquid at temperatures across what a CPU would see, unless you took your powered down laptop to Antarctica or Siberia and left it out in the cold where the galinstan would solidify and expand by about 3% (potentially tearing the heat sink off of the CPU if you did not make room for expansion).

    Gallium loves to alloy with other metals; even with copper the metal ions will migrate in to the copper metal, gradually creating a copper-gallium alloy that is grey-silverish in color.

    What makes this happen much faster is the electrochemical potential of the two metals. This is the same principle that makes an old fashioned lead-zinc battery function. Two metals in contact create an electrical potential and current will flow. One will oxidize (corrode) and the ions will move other to the other metal where there is a reduction (or plating) effect. Electrochemically speaking, the most negative metal will sacrifice itself by corroding away to build up a layer on top of the positive metal.

    Electrochemically speaking, gallium has a potential of -0.53 volts and copper has a potential of +0.334 volts. The difference between the two metals is going to create a "battery" with a potential of 0.864 volts.

    Gradually this battery will deplete; as the gallium migrates over to the copper. You may have heard some stories about how gallium "dries up" and evaporates. In many of those cases it is likely that it acted like a battery and migrated over to and became part of the copper heat sink.

    If you take apart the heat sink from the CPU and clean it up you may find that the copper heat sink is colored a silverish-grey that resists efforts to even buff it off with a scrubbing pad. The stuff you can polish off is a corrosive residue of oxidized gallium and the stuff you cannot remove easily is now an alloy.

    One good piece of news is that if this happens repeatedly the battery action will stop, as it is pretty inefficient to put gallium and have it migrate over to a gallium-copper alloy.

    There might be a few things that are not so good news. It is possible that the gallium-copper alloy is much less a thermal conductor than pure copper. It is also possible that the alloy is not as structurally sound as pure copper or becomes brittle (gallium is pretty brittle when it is solidified). You may not even notice the first instance of less thermal conductivity unless you were in a laboratory setting and counting calories (thermal) and heat transfer characteristics, the actual difference may only be 1-2 C in a computer and that can get lost in all of the other variables of how computer CPU's cool, loading, power, air temperature, dirt on the fins or fan inefficiencies.

    The second thing regarding brittleness might only be a concern on a vapor-phase heat sink where the metal is very thin. Again, this is not proven by any particular instance but there is always the potential of the vapor phase heat sink of fracturing and letting out the magic cooling juice, right on to the CPU. (do not panic over this statement, I am not backing it up with any examples but maybe heat sink manufacturers should look in to it).

    If you look at the IHS (integrated heat spreader) on top of a CPU it is made of nickel plated copper. If you ask the question "why did they put nickel on top of copper? copper is so much better of a heat sink material) it may be due to the corrosion vulnerability of copper.

    Actually having the IHS plated with nickel is a good thing, it has a galvanic potential of -0.3 volts, much closer to liquid metal so the corrosion between the CPU IHS and the liquid metal is almost nil.

    What might be an idea is to electroplate the contact area on the heat sink with nickel (nickel can electroplate directly to copper) and it would be over an area about the size of a postage stamp. In manufacturing that plating takes about five minutes but there is quite a bit of set-up to keep the nickel just where you want it and not over the entire heat sink.

    You "could" electroplate nickel at home but it is a fairly toxic process and generates hazardous waste.

    Again, maybe heat sink manufacturers can consider some sort of plating process.

    Another alternative is a barrier that prevents electrical connection (that allows the battery action of galvanic corrosion from happening) but still has very high thermal conductivity. We have had this with regular ceramic, alumina (and even silver based) thermal pastes. No matter that the word aluminum or silver is in the name, most are pretty horrible electrical conductors.

    This problem is pretty unique to just the liquid metal TIM (thermal interface materials) crowd but it is something that we should be aware of.

    (BTW, I nickel plate and then silver plate my heat sinks but I have the experience from doing this with jewelry. It is an expensive and potentially toxic thing to get in to just for a heat sink).
     
    bennyg, FrozenLord, Arondel and 18 others like this.
  2. kosti

    kosti Notebook Virtuoso

    Reputations:
    595
    Messages:
    2,163
    Likes Received:
    464
    Trophy Points:
    101
    First off, very informative post.

    Question:

    Have you had any experience with liquid metal paste actually staining a silicon die?

    I applied some Thermal Grizzly Conductonaut on my 920xm. After a couple of months, temps started shooting up dramatically with low CPU usage. I removed the heatsink and found the paste had hardened almost completely, except for a couple of tiny spherical clumps of paste here and there. I noticed the stain left behind on the copper heatsink (which I had to sand down with 800 grit sandpaper to remove) but I also noticed that it left a film on the silicon die as well. I tried removing it with different solvents (even acetone) but nothing removed it. I ended up using a scotch-brite pad to get it off and that obviously that left some fine scratches on the die but it doesn't seem to affect performance.

    I am now using Gelid GC Extreme which is a very good non-metal paste. I'm pretty hesitant about using the liquid metal stuff again because of the staining, but mostly because it hardened so soon. The temp drops were initially phenomenal however and really helped me with overclocking.
     
    Vasudev, hmscott and Dr. AMK like this.
  3. Tishers

    Tishers Notebook Consultant

    Reputations:
    90
    Messages:
    105
    Likes Received:
    196
    Trophy Points:
    56
    I do not know if it is necessarily bad for the stain to exist or even if there is a migration of some of the metallic elements in to either the IHD, heat sink or the ceramic at the top of the CPU die. It is just an observation and there has been some research done in to how gallium (and its alloys like galistan) interact with copper in much more critical applications than a computer heat-sink. We know that aluminum heat sinks would be a awful thing.

    ( gallium and aluminum reaction: )

    ( research paper on gallium and galinstan alloy compatibilities with copper: (you will really need to be a nerd to read this) http://www.dtic.mil/dtic/tr/fulltext/u2/a201304.pdf )

    Copper and nickel will stain and have some minor pitting, leaving behind a "sludge" (their words in the research paper) but if you replenish the LM the generation of sludge slows down; they did not note a complete breakdown of copper. Additional investigations were in to the use of galinstan as an electrical conductor in homopolar motors (look it up, quite cool) and as a thermal conductor in future fusion reactors.
     
    kosti and Dr. AMK like this.
  4. LTBonham

    LTBonham Notebook Evangelist

    Reputations:
    181
    Messages:
    350
    Likes Received:
    52
    Trophy Points:
    41
    Interesting posts. I have been using CLU for about 18 months now, before that Gelid GC Extreme.

    My latest application lasted 1 yr and I finally started noticing that temps were starting to creep up again. After another repast, it seems as good as I can remember even with the copper staining. If it does indeed cause issues with Copper, I haven't run into it yet.

    If it does ever cause issues, I will share on the forum.
     
    Papusan, kosti and hmscott like this.
  5. triturbo

    triturbo Long live 16:10 and MXM-B

    Reputations:
    1,565
    Messages:
    3,733
    Likes Received:
    1,184
    Trophy Points:
    231
    @Tishers - Thanks for the post. A question - any chance it happens at a certain temperature? I think that once the CPU passes ~80ºC (maybe a bit more) it's a steep downhill from there, it degrades pretty fast. This is my observation at least.
     
    hmscott likes this.
  6. Tishers

    Tishers Notebook Consultant

    Reputations:
    90
    Messages:
    105
    Likes Received:
    196
    Trophy Points:
    56
    Most assuredly any sort of chemical reaction will happen quicker as temperatures increase. In the reading I have done on the topic the threshold seems to be up around 300 C before things get really out of hand.

    That is one of the reasons on why I brought up the galvanic reaction between dissimilar metals. If two different metals are in contact and you create another electrical path around them you get current flow. It is exactly the same principles that make up a "battery". One will sacrifice itself and corrode away, while the other metal accumulates a crust of oxides and electrolytes.

    If there is a galvanic reaction you can easily check it. Take a piece of Nickel (even a real us Nickel), a dab of TIM (galinstan) and a piece of copper (a penny, but that is only a copper coating). If you have a voltmeter you can measure the electrical potential from the penny to the nickel, with the galinstan acting as a metal-electrolyte in between. The value will be in the fraction of a volt.

    Now imagine that as your computer heat sink on a CPU. The heat sink may be electrically isolated (by a plastic bracket) or electrically "bonded" with screws to metal that are running in to the motherboard where they are part of the ground-plane (definitely if they are going in to a steel chassis). With the electrically bonded connection there will be current flow, just like discharging a battery.

    The old zinc-carbon batteries basically corroded away the cathode that is the outer case of zinc (the negative most connection) to "protect" the carbon anode in the battery (the positive most connection). That is why dead batteries are never a good thing to leave in an item, they corrode away and ruin the item. But you also know that even charged batteries can corrode away all by themselves over time (albeit much more slowly). There are enough chemical impurities in the battery that it has its own self-leakage and even things like how much a metal is tempered (like a die extruded heat sink is harder and softer in some areas) will cause small circulating currents that cause localized pitting.

    It has probably been entirely overlooked by many in the overclocking/ liquid metal TIM crowd but corrosion is a very real concern. You can bet that Intel has thought of it and it would not be surprising at all to find that they have researched the long term implications of liquid metals CPU package.

    I am aware of it because of one of the many hats I have worn in my career was as a corrosion engineer for an oil company. Dissimilar metals were always a concern and if we made a mistake in protecting and bonding it could result in a spill or an explosion.
     
    KY_BULLET, Papusan, t456 and 2 others like this.
  7. Raidriar

    Raidriar Notebook Prophet

    Reputations:
    1,133
    Messages:
    4,559
    Likes Received:
    2,890
    Trophy Points:
    231
    That's pretty wild, and it makes perfect chemical sense. I say I haven't had issue with CLU on my M18x R2 for the past 2 years, but the number of extended heat/cool cycles it has been exposed to is very, very minimal, so to say liquid metal does have a "timer" in a sense for how long the system has for the gallium to alloy itself to the copper. Pretty amazing

    can be mitigated by adding more liquid ultra/conductonaut as you said, but it would reduce thermal conductivity as it is transferring to the gallium-copper alloy instead of pure copper instead.

    mind blown. I wonder if CL or ThermalGrizzly know of this process..
     
  8. triturbo

    triturbo Long live 16:10 and MXM-B

    Reputations:
    1,565
    Messages:
    3,733
    Likes Received:
    1,184
    Trophy Points:
    231
    @Raidriar - Have you ever passed 80~85ºC? I think that's what "activates" it. As long as you keep under 80, I think everything is good (for prolonged use, as I said, once certain temp is passed it degrades in a matter of weeks, not to say days).
     
  9. Raidriar

    Raidriar Notebook Prophet

    Reputations:
    1,133
    Messages:
    4,559
    Likes Received:
    2,890
    Trophy Points:
    231
    oh yes, things get very hot at 4.4ghz full load 4 core. usually tops out around 84 degrees
     
    jaug1337, Papusan and triturbo like this.
  10. jclausius

    jclausius Notebook Virtuoso

    Reputations:
    983
    Messages:
    2,492
    Likes Received:
    989
    Trophy Points:
    131
    I've seen others talk about this (@Mr. Fox - I believe), and most posts say that the staining is normal, and can be left behind as it won't affect thermal transfer into the heat sink.

    From the posts I've seen about this, if there is a poor fit between the heat sink and target component, a small amount of air may be getting to the Liquid Metal causing it to harden up over time. Some have suggested entirely surrounding the liquid metal with something like GC Extreme in order to prevent any extra air to the LM in cases like this.
     
    Papusan and Mr. Fox like this.
Loading...

Share This Page