Galvanic Corrosion for muppets: Impress your friends!

[charliek]

Whether you are interested in water cooling your PC, riveting together your DIY nuclear sub, or just want something different to chat about in the bar, you ought to know about Galvanic Corrosion. You may not think that you do, but you do.

People kept mentioning it to me, and I hate it when other people seem to know more than I do - so I've been doing some reading up. Here, for you delight and entertainment is what I've found out:

(Please bear in mind that this isn’t supposed to be a detailed technical discussion that would hold up at a marine engineer’s conference – it’s just meant to hold up at LAN parties and dinner parties. If you notice any glaring mistakes, then please have a quiet word with me at charlie@stopthatitssilly.com, or on the private message system here)


What is Galvanic Corrosion, and why is it a bad thing?

It isn’t necessarily a bad thing. The first ever batteries worked on the principal of Galvanic Corrosion: If you put two dissimilar metals (like steel and aluminium) in an electrolyte (a liquid that conducts electricity, like water), a volt meter will show you that a current flows between them.

The first ever batteries were made by sandwiching layers of different metals with salt water soaked stuff – the so called “Wet Cell”. (Salt water is *seriously* electrolytic, for you nuke sub folks)

Unfortunately, you don’t get free electricity this way, there is a cost. The cost is corrosion. The less ‘noble’ metal (further up in the galvanic series, below) will corrode quicker than normal, and the more ‘noble’ metal (further down in the galvanic series, below) will corrode more slowly than it normally would.

Note the phrase ‘normally would’ – that’s because many metals, like iron, will corrode quite happily all on their own when immersed in certain liquids, like water.


Why might it affect me?

Well, if you are water cooling a computer, the chances are that you’ll be making a loop of electrolyte (water) in which are immersed several metal things (inside of radiator, inside of CPU block, inside of GPU block, inside of chipset block…). I trust I don’t have to draw pictures?

If you are not careful, you might easily end up with two metals that are far apart on the Galvanic Scale (below). Copper, Bronze, Brass and Aluminium all spring to mind.

Let’s say that you had a copper heat block, and a stainless steel reservoir and radiator. Copper is more anodic (less noble) than Stainless Steel (more noble). The galvanic corrosion effect would mean that your copper heat block would corrode faster than normal. The Stainless Steel would be protected and would corrode less quickly.

Having our copper block corroded is a Bad Thing™ (especially if it’s full of water and inside our computer)


What do I do to avoid Galvanic Corrosion?

Stick to a single rung on the Galvanic ladder.

Metals at the same level on the Galvanic Scale will not cause Galvanic Corrosion between themselves. Even if they are not on the same level, the degree to which the anode’s corrosion is accelerated increases the further apart the metals are on the scale.

If you put Magnesium and Platinum in an electrolyte, the Magnesium will corrode much faster than if the combination were Magnesium and Aluminium.

Surface Areas

The relative surface areas of the metals are important

If the metal that’s being corroded (the anode) has a small surface area and the metal that’s being protected (the cathode) has a large surface area, then the corrosion will be particularly quick. An aluminium bolt in a large stainless steel plate will corrode rapidly.

If it’s the other way around, however, (a small copper heat block connected to a large aluminium reservoir, for example) then the corrosion of the anode will be much slower.

Sacrificial Lamb

There’s a technique referred to as ‘sacrificial anode protection’ whereby a lump of even more anodic metal is added to the system, which will then corrode instead of the important bits.

For various reasons too complicated for this discussion, Zinc is a popular choice (maybe it’s because magnesium has an embarrassing habit of burning fiercely with little provocation). It’s towards the most anodic end of the Galvanic Scale.

Zinc blocks were attached to the hulls of ships to protect the rivets. The blocks could then easily be replaced without having to dismantle the ship.

Sacrificial anodes can also be seen in domestic water heating systems. An interesting article explaining the use, and showing a 'before and after' of a sacrificial anode, can be seen here

The sacrificial anode is connected by a wire to the metal that it serves to protect – in order to establish a ‘path of least resistance’ – this ensures that it will always be ‘eaten’ in preference to the bit that would otherwise have been the anode. In effect it’s forcing it to be a protected ‘cathode’ instead of being the ‘anode’ - this means that there must be an electrical conductivity path between the sacrificial anode, and the cathodic metal that it serves to protect, as well as the electrolytic path provided by their being in the same body of water (electrolyte).

Make Up Powder and Paint

Many coatings are available that will prevent, or at least slow, Galvanic Corrosion.

Some of them contain ionic substances that will function like sacrificial anodes. Ships, for example, are often painted with zinc-based paint that has to be replaced every so often.

Others coat the surface of the metal with an inert layer that effectively separates the metal from the electrolyte. Anodised aluminium is a good example of this technique.

A word of warning though: if you protect your anodic metal (by painting it, for example) and a small area of the paint is damaged, then you are left with an exposed anodic metal of much smaller surface area than the cathode, which means that it will corrode particularly quickly at the site of the coating damage (see Surface Areas above).

Stop The Electrolyte electrolyting

Another means of preventing the process of Galvanic Corrosion is by preventing the electrolyte from being electrolytic any more. De-ionising water is one example, unfortunately the fact that it’s running in contact with metal will soon re-ionise it, so it needs regular replacement.

The addition of ‘corrosion inhibitors’ does a similar thing, by keeping the ions in the water occupied, so they can’t be eroding the cathode. Again, because this process actually consumes the corrosion inhibitor, it needs topping up periodically.


Conclusion

The chemistry of the water and the metals in your water cooling loop means that you run the risk of causing, or accelerating, the corrosion of one or more of its components.

In order to avoid, or minimize the risk of accellerated corrosion of something important there are some things you can do:

- Keep all the metals in the system the same, or as close on the Galvanic Scale, as possible.

- If you have to mix metals, try to make the anode’s surface area much bigger than the cathode’s.

- Consider using an even more anodic metal as a ‘sacrificial’ anode, say by plopping a lump of zinc into the reservoir every now and then. Connecting it to the anodic metal that you want to preserve by a wire may help too.

- Inert coatings, such as that found on anodized aluminium, effectively block the galvanic corrosion process or, in the case of zinc based marine paints, act as sacrificial anodes themselves.

- Consider the cooling liquid itself: De-ionised water inhibits galvanic corrosion for a while, as do corrosion inhibitors such as those found in some anti-freeze and coolant additives. In both cases the effect decays, and the medium needs replacing periodically. Bear in mind, however, that if you use a completely non-conductive liquid, then you will lose any benefits that a sacrificial anode would have in delaying normal corrosion.

Finally, in the case of water-cooling a domestic computer, we need to make sure that our solution is safe, easy to maintain and cheap. (Otherwise we could just coat the whole thing in gold, using gold arsenide which is onscenely poisonous, and have done with it. Either that, or commission a Cray, and let it be Cray’s problem)

Appendix 1: The Galvanic Corrosion Series.

Here’s that series in all it’s glory. Well, not all it’s glory, actually, because this is an extract, a more complete list is at http://www.mcnallyinstitute.com/CDweb/g-html/g001.htm or at http://www.ocean.udel.edu/seagrant/publications/corrosion.html

This is the ‘Least Noble’, or ‘Anodic’, end

MAGNESIUM
ZINC
ALUMINUM
MILD STEEL, WROUGHT IRON
CAST IRON, LOW ALLOY HIGH STRENGTH STEEL
STAINLESS STEEL
ALUMINUM BRONZE
LEAD - TIN SOLDERS
LEAD
TIN
NICKEL (ACTIVE (note 1))
BRASSES
COPPER
SILICON BRONZE
NICKEL SILVER
STAINLESS STEEL (note 2)
NICKEL, ALUMINUM, BRONZE (CA 630, 632)
SILVER SOLDER
NICKEL (PASSIVE (note 1))
STAINLESS STEEL (PASSIVE (note 1)) (note 3)
NICKEL - MOLYBDEUM - CHROMIUM - IRON ALLOY (PASSIVE (note 1))
SILVER
TITANIUM (PASSIVE (note 1))
ZIRCONIUM
GOLD
PLATINUM

This is the ‘Most Noble’, or ‘Cathodic’, end.

(note 1) Some metals change their level of galvanic potential depending on whether they are just sat in a bath of electrolyte (Passive), or are (for example) in fast flowing seawater, being lived on by algae (Active)
(note 2) Some alloys of stainless steel
(note 3) Other alloys of stainless steel

Appendix 2: How can I ever thank you enough for this fascinating masterwork?

All part of the service.


Appendix 3: References

http://www.ocean.udel.edu/seagrant/publications/corrosion.html
http://www.mcnallyinstitute.com/CDweb/g-html/g002.htm
http://www.finishing.com/266/26.html
http://www.capsante.com/Articles/howto_sacrificial_zincs.htm

And, of course, dear old Google.

Edit: Notes in Appendix 1, and comment re. non-conductive liquids.
Edit: Sacrifical anodes need an electrical path to the cathode the serve to protect, as well as being in the same electrolyte (water), see 'Sacrifical Lamb', above.
Edit: Added a reference.
Edit: Added a bit about the dangers of scratched coatings on anodic metals to 'Lipstick Powder and Paint' above.[/url]
Edit: Added reference to sacrificial anodes in domestic water heating (thanks 'drownmypc')
_________________
Charlie

[sthayashi]

Of all the things I've ever read on this site, I think the above article has been the most informative. It deserves to be stickied.
_________________
My Power Rig, Storage Rig, HTPC and Main Rig

[Edward Ng]

Would using a nonconductive coolant, such as Fluid XP+, solve this problem?
_________________
Contributing Writer, SPCR
NgTechnik
Want something reviewed? Help us get samples!
Donate for Patron or Friend Status!

[Rusty075]

Stickied.

Any significant corrections/additions should be edited into Charliek's first post, so that people will get the best information at the very top of the thread.


I'll add a minor amendment:

Most of what WC'ers online describe as, and show pictures claiming to be galvanic corrosion, isn't. It's just plain oxidation. If your copper just turns green...its not galvanic corrosion, its just plain old corrosion.
_________________
Senior Contributing Writer, SPCR

[charliek]

Oooh, you're all too kind

I'm generally greatful when someone goes to similar effort for me and I was going to do the research anyway, so...

Ed - I don't know the product you mention, but if it's electrically non-conductive, then yes it'll prevent galvanic corrosion, as for regular corrosion (rust, copper verdigris, etc...) its lack of conductivity won't make any difference to that. Its other chemical properties might, in either direction. Which ties in with:

Rusty - exactly: Galvanic corrosion effect either speeds up, or slows down, existing corrosion effects. However the good news is that the 'sacrificial anode' method of galvanically protecting your other metals will effectively slow down corrosion on all of them, except for the sacrifical anode (assuming you chose one that's more anodic than the rest - Zinc for us mortals).

Next time I bump into a chemist or a physicist at a party (it happens you know), I'll ask him or her to look this up and down and comment. (unless, of course, she's a 'her' and there are other things that I decide to discuss with her)
_________________
Charlie

[peteamer]

Well done Sir.

[HammerSandwich]

Nicely researched; well written!

I didn't notice any mention of eliminating the electrical connection between the anode and cathode. This should cause a substantial reduction in galvanic action.

[Gooserider]

Difficlt to do HS... The anode and cathode are connected by the conductive coolant... The coolant is even arranged in a loop so you have a potentially complete circuit!

Overall a good article, I would add however that anti-corrosives, coatings, etc. in part because of the wearing out issue, won't completely stop GC, but merely slow it down. Now if one slows it down enough that the hardware lasts longer than the effective service life of the PC, it could be argued that this is "good enough", but I tend not to buy that stance since 1. I usually keep my PC's far longer than most folks, and 2. WC hardware tends to be more 'reusable' than other PC components so you might well find yourself wanting to use WC hardware long after the PC it was originally installed on has been retired.

Gooserider
_________________
Building Dual Athlon MP system, Tyan mobo, U160 drives, Server Cube case, Linux ONLY, lots of other goodies. Will water cool, attempting to make as silent as possible.

[HammerSandwich]

[charliek]

[Edward Ng]

So it would be beneficial to, say, stick in a piece of zinc to replace a piece of tubing somewhere in the loop...

Anybody know of a plce that would custom fab such a piece, online, and send it out?
_________________
Contributing Writer, SPCR
NgTechnik
Want something reviewed? Help us get samples!
Donate for Patron or Friend Status!

[charliek]

[Edward Ng]

Hmm...

Any ideas on how to do this without a reservoir? I'm looking to avoid a reservoir as a closed circuit with no standing water maintains full momentum and allows the pumps more efficient operation. This is why I was thinking of substituting a portion of the circuit tubing for zinc piping.
_________________
Contributing Writer, SPCR
NgTechnik
Want something reviewed? Help us get samples!
Donate for Patron or Friend Status!

[charliek]

[HammerSandwich]

If the wire between sacrificial anode and cathode is essential, then I'd bet that the loop CANNOT be completed by the electrolyte (and requires a separate conductor).

As far as sacrificial anodes go, I'd be concerned about chunks of zinc breaking off and clogging the loop. The severity of this will depend on the block and a million other things, but I don't see a lot of potential for quiet pumps and effective filters.

Anyway, check out the bottom of McMaster page 2108.

[charliek]

[HammerSandwich]

[Gooserider]

Just as a hypothetical to add to the 'is there a loop?' question... I do NOT know this to be factually true, and it might not be a factor, however think about the notion that we are always talking about loops and circuits when we are discussing a WC setup. I think it is quite reasonable to assume that one has two current paths in a WC setup, namely the waterline going TO the part and the line coming FROM the part. In addition we have a nice circulation of water to carry the corrosion products from one part to the next.

A way to test this might be to measure the ionic content of the coolant in different parts of the loop - if my theory is correct you would see a different ion mix (and polarity?) in the Al -> Cu section than you would in the Cu -> Al section.

Just a thought, I could easily be in error.

Gooserider
_________________
Building Dual Athlon MP system, Tyan mobo, U160 drives, Server Cube case, Linux ONLY, lots of other goodies. Will water cool, attempting to make as silent as possible.

[charliek]

[Gooserider]

You could well be right Charlie. My theory is based on the idea that since you have a flow of coolant containing the particles put out by each source, the different blocks would in effect act as filters for the other block's particles. Sort of like if one substituted a pumped glass of lemonade in the example you cited - I would expect the current to carry the ions around the circle, and each block would suck up what was attracted to it, and spit out what wasn't so the ratios between the cation and anion would vary depending on where you measured them in the loop. (Not that it matters a great deal, or that I have the sort of equipment needed to measure the difference...)

Gooserider
_________________
Building Dual Athlon MP system, Tyan mobo, U160 drives, Server Cube case, Linux ONLY, lots of other goodies. Will water cool, attempting to make as silent as possible.

[Bat]

No, no, no.

In those lemon things, at the surface of the zinc or aluminium or whatever the more reactive metal is, atoms lose electrons to form positively charged metal ions (cations) in solution.

The electrons flow through the wire and clock and more wire to the other electrode. There, at the surface of the copper electrode, they react with something in the lemon juice or potato juice.

In principle it might be with oxygen and water to give negatively charged hydroxide ions (anions), or it might be with some positively charged ions (cations) to give something else, e.g. turning iron triple-positive to iron double-positive ions.

N.B. There is an electrical circuit involved here.

Also note that the overall reaction is between the zinc (which loses electrons) and some oxidising agent in solution (which gains them). The circuit and copper just make the process faster.

In a water-cooling system, the only candidate for the oxidising agent is oxygen from the air. So, when filling your system use water which has been boiled for a while to drive out the oxygen.

Finally: the shape of the water (loop or not) makes no difference. Gooserider, just as pumping the water keeps the temperature almost constant at all points, it will do the same for the concentrations of the various solutes.

[Bat]

I was reading about that "FluidXP" stuff and it's mostly water, so I'm not impressed.

[echoes]

[TheAtomicKid]

Unless I'm mistaken, and I may be, chemistry isn't my strong suit, having managed to somehow bypass it in highschool...

All that zinc has to go somewhere... it doesnt just 'disappear into the ether'... my guess is it's gonna get deposited onto your important bits of metal... how is THAT going to affect your heat transfer?

IMO, best solution... make every bit of metal in your watercooling loop, the same metal.. preferably the same grade... and use inhibitors. Replacing your liquid supply/refreshing the inhibitors frequently is probably a real good idea too.

Not that I'm against watercooling.. I'm certainly not.. been pondering it myself lately but.. all this is certainly food for thought, for staying with air cooled solutions.

VERY informative article, thankyou. *goes off to think some more*

TAK

[Butcher]

The zinc goes somewhere - it becomes zinc oxide and yes it'll get deposited all over everything.

[cercle]

Charliek+others:nice sharing of knowledge,and then,... is there a possibility for spcr to test/teach us Watercooling "KNOWHOW " :listing copper blocks family,silent copper pumps..... A french site tested pumps,PCSILENCIEUX,i think. Spcr may enlarge its specialty,including more water to its air

[Olle P]

I agree with those stating that keeping the different metals insulated from each other (not counting the electrolyte) goes a long way in slowing down galvanic corrosion.
Just think of the wet cell batteries mentioned: As long as you don't connect the electrodes to any external circuit they will erode very slowly.

In water cooling, the main target for galvanic corrosion can be expected to be the joints where a cooling block (or radiator) of some metal (usually copper or aluminium) is attached to barbs/fittings of some other metal (usually brass or stainless steel). There you have a direct metal to metal contact as well as the electrolytic bridge!

Cheers
Olle

[RBBOT]

An interesting read. Does anyone know exactly what the magnitude of difference galvanic corrosion makes to the overall corrosion rate?

e.g. Say for example we had a piece of aluminium of dimensions such that if put in water on its own 10% of it would corrode in a given time period, and we have a piece of copper that did the same, which would probably be of a different size.

If we put them in the water together how far is galvanic corrosion going to shift the 10%:10% balance in corrosion rates?

[fri2219]