Metal powder for heat sinking

[bootlegin]

Full disclaimer. this is just a hypothetical. I don't plan on doing this to something I care about. Also this idea is only for heat sinks that clamp or screw down onto the board.

I noticed that metal based thermal pastes have an incredibly low thermal conductivity compared to their pure metal counterparts, so ignoring the electrical conductivity and possibility of making a mess, what if straight metal powder like copper or silver were used to fill the voids between the CPU die and heat sink instead of thermal paste?

In such a scenario, both the die and heat sink block would be sanded down flat to remove the concavity, then polished to about 1000 grit. Then afterwards, the polished faces would be vigorously rubbed into a thin pile of sub-micron silver or copper powder against another flat surface to fill in any voids (I personally believe that simply rubbing the powder in should squeeze the particles into the voids and hold). and finally, the CPU would be installed, a small pile of powder would be placed in the center of the CPU, then the heat sink would be slowly spun left and right in small arcs using the center axis while very carefully lowering (to prevent the powder from blowing out and dispersing in the air) to evenly distribute the powder and fill any remaining voids. And finally, once powder starts falling out the sides, clamp the heat sink down onto the board for a tight connection.

[tooki]

I’m thinking you don’t understand how thermal paste works.

You start with smooth surfaces to begin with. But even a “smooth” surface has microscopic imperfections. So when you press two such surfaces together without any paste, you get contact where high spots touch high spots, and have gaps elsewhere. In those gaps you have air. Air is a really bad conductor of heat. The idea behind thermal paste is to replace those air gaps with something that’s a better thermal conductor than air.

What’s in thermal paste? You have fine particles of a highly thermally conductive material (like metal oxides) in a silicone or petroleum grease.

If you’re lucky, a given gap will be perfectly filled by a fine particle. Or multiple fine particles will be in direct contact with each other, creating a chain of thermal conductivity. If not, you have some microscopic layer of grease between a particle, another particle, and/or one of the surfaces. So basically, you have highly thermally conductive materials (surfaces and particles) with grease in between.

Your idea of metal powder is… fine particles. But what fills the gaps between those particles? Air.

Which is the better thermal conductor, air or grease?

And that’s why we embed the thermally conductive particles in grease.

[bootlegin]

Alright gotcha this makes a lot of sense. My next question then is why are thermal pastes so bad at conducting heat compared to their pure metal counterparts? I mean even the most conductive paste I can find, boron nitride paste, only has a conductivity of a little over 30 W/m.K whereas CBN has 1300 W/m.K. Is the ratio of metal particle to grease really low? Would a larger amount of metal particles increase the conductivity?

[tooki]

Would a larger amount of metal particles increase the conductivity?

Up to a point, yes.

Is the ratio of metal particle to grease really low?

No. Thermal pastes are already >70% solids by volume, and at some point it’d cease to be a paste and would become a putty that doesn’t flow and could lead to air inclusions.

Alright gotcha this makes a lot of sense. My next question then is why are thermal pastes so bad at conducting heat compared to their pure metal counterparts? I mean even the most conductive paste I can find, boron nitride paste, only has a conductivity of a little over 30 W/m.K whereas CBN has 1300 W/m.K.

Heat conduction requires physical contact. Thermal resistance behaves just like electrical resistance: a narrow conductor is a worse conductor of heat or electricity than a wide one, for any given material. So even if your substance is 75% solid, if the physical shapes mean lots of bottlenecks, that increases resistance.

For the sake of illustration, let’s pretend our theoretical paste has metal particles that are little spheres, like ball bearings. The three-dimensional packing density of equal spheres is about 74%. Now look at the contact area between any two spheres: it’s a theoretically infinitesimally small contact patch. That tiny patch is bottleneck, an area of high resistance (thermal or electrical). We have a bunch of spheres, so with many of these contact patches in parallel, in aggregate we do get some useful conductivity, but it’s far, far less than if you took those spheres and melted them down into a block.

I have no idea what the actual particle shapes in thermal greases are, nor what those shapes’ packing behaviors are. But you certainly can’t get anything approaching conductivity in proportion to the solids % in the paste (by which I mean that e.g. 75% aluminum particles in the paste won’t get you anywhere near 75% of the conductivity of solid aluminum.)

[bootlegin]

What binder/grease is typically used in a thermal paste? If a more thermally conductive binder were to be used instead of what's commonly used in thermal paste, would that help increase the overall thermal conductivity?

[Siwastaja]

My next question then is why are thermal pastes so bad at conducting heat compared to their pure metal counterparts?

Because they are good enough. Because they are used in very thin layers, even the modest conductivity is fine; you get diminishing returns on improving it even further. There are other factors, like being electrical insulator, being cheap enough to manufacture, non-toxic etc. etc. There is no point in sacrificing any of these to increase thermal conductivity to a point where it decreases your CPU temperature by 0.5 degrees.

[tooki]

What binder/grease is typically used in a thermal paste?

Ahem:

What’s in thermal paste? You have fine particles of a highly thermally conductive material (like metal oxides) in a silicone or petroleum grease.

You could also, you know, do some research of your own. Read the technical datasheets for different thermal greases and adhesives. Look at what’s out there and look at what’s in it.

If a more thermally conductive binder were to be used instead of what's commonly used in thermal paste, would that help increase the overall thermal conductivity?

Such as…?



What makes you think the manufacturers of thermal pastes wouldn’t have researched this already?



Here’s a really good thing to do when doing this kind of mental exercise, thinking about how one might improve a product: rather than wondering “how about doing X?”, ask “why don’t we do X already?” Searching for that will very often turn up evidence that someone else has thought of it before, and looking into that often reveals why that avenue failed. There are usually good reasons why we don’t do something “obvious”.

[Kleinstein]

Thermal conduction gets 2 extra complications: if a material is very thin (e.g. less than about  1 µm or a fraction of the wavelength of thermal IR) there can be additional IR type couling between the 2 sides that is higher than normal due "near field coupling".  The interface between different materials can add extra thermal resistance (Kupica effect), if the 2 materials have different ways of conducting the heat (e.g. metal to grease is such a candidate). So super fine particles may be controproductive adding more interfaces.