Thermally Conductive, Electrically Insulating, Thin, Corrosion/Alloy/Permeability Resistant Conformal Coatings?

[Daharen]

Doubt this exists, but if there are any material engineers in these forums, is there a conformal coating available that would be practical to apply to a die after lapping it to prevent liquid metal from permeating the silicon and causing damage to the underlying circuitry, without significantly inhibiting the thermal gains afforded by lapping the die in the first place, causing the die to cease to be level, or creating any risk of electrical conduction (Goes without saying with a conformal coating, but I know there are different grades for different uses)? 

I feel like if this was possible it would already have been done or mentioned somewhere, but then again I could be wrong, as it could be the case that most people just don't want to lap their dies that far down, and thus never had to deal with this issue. If there is a solution, I would be immensely interested in what it is and where I can purchase it. If it's an extremely expensive industrial compound out of consumer price ranges, I'd still be interested, though I can't see spending more than $500.00 for what is really not a significant benefit, so it would just be curiosity at that point. 

[RejZoR]

You're not lapping a die. And die already has a protective layer on top which doesn't allow liquid metal to get in.

[WoodenMarker]

The protective layer including more diamond in it may increase thermal conductivity but would probably be less reliable. https://aip.scitation.org/doi/pdf/10.1063/1.4978043

I'm not sure what the cost would be but it's not within your budget nor practical. It's a process that's most likely limited to custom orders of silicon wafers and the process for getting it on a specific cpu would be proprietary to its manufacturer. 

[Daharen]

4 minutes ago, WoodenMarker said:

The protective layer including more diamond in it may increase thermal conductivity but would probably be less reliable. https://aip.scitation.org/doi/pdf/10.1063/1.4978043

I'm not sure what the cost would be but it's not within your budget nor practical. It's a process that's most likely limited to custom orders of silicon wafers and the process for getting it on a specific cpu would be proprietary to its manufacturer. 

Yep, that's what I was looking for, and I somewhat figured. I have seen people lap dies and put LM on it, but I've heard of no long term evaluations of this process to determine if it eventually does damage the CPU. 

I am curious whether the LM will continue to bond indefinitely over time, and slowly alloy down to the circuitry if one waits long enough, or if its a process that naturally tends to dissipate the further from the source of the original substance the infiltration occurs, in which case eventually assuming you leave enough silicon on the chip, it will stop, and you'll have little to no risk of causing further damage. I know usually this sort of chemistry is dependent on the molecular arrangement differences between the alloy and the original substances as well as the chemical potential from their bonding, but I have no idea how to figure out for myself what is most likely to happen without testing it, and no one seems to have felt the need to lap their die and expose it LM for a long period of time (I know Der8aur did put LM on a lapped CPU with direct die cooling, but I don't think he left it that way and tested it over any significant period to see if it caused damage). 

Anyway, it might be something I test out when I upgrade, I'll make my 8700k a a guinea pig for the experiment, and see how long it takes to fry itself, or if it lasts indefinitely. Should be fun to find out. I'll just scrape off a VERY thin layer with a mild lap, just to get rid of the barrier, but leave as much silicon intact as possible. If it turns out to be practical, and lasts awhile, Ill cut the CPU in half after say two years, measure the depth of the alloying, and use it to gauge a safe amount to lap any given die, after looking up the specs for the size of the silicon thermal interface they provide. 

Naturally if it does kill the chip, then there's no point in doing this, and CPU lapping becomes pretty pointless (Unless you have a REALLY uneven chip, in which case you are pretty much shit out of luck, and must use thermal paste from that point on if LM does slowly and indefinitely leech into the silicon). 

[WoodenMarker]

15 hours ago, Daharen said:

Yep, that's what I was looking for, and I somewhat figured. I have seen people lap dies and put LM on it, but I've heard of no long term evaluations of this process to determine if it eventually does damage the CPU. 

I am curious whether the LM will continue to bond indefinitely over time, and slowly alloy down to the circuitry if one waits long enough, or if its a process that naturally tends to dissipate the further from the source of the original substance the infiltration occurs, in which case eventually assuming you leave enough silicon on the chip, it will stop, and you'll have little to no risk of causing further damage.

It still seems to be working as of 5 months ago. https://www.youtube.com/watch?v=tnd2LO0IBic

 

I'm guessing you'd replace the cpu before it becomes an actual problem. It would depend on how much of the original coating is present.

[Daharen]

On 10/4/2019 at 5:31 PM, WoodenMarker said:

It still seems to be working as of 5 months ago. https://www.youtube.com/watch?v=tnd2LO0IBic

 

I'm guessing you'd replace the cpu before it becomes an actual problem. It would depend on how much of the original coating is present.

That's awesome and great to know. Thanks for posting this :).