Secret file thermal paste: Increasing layer thickness at high temperatures, ageing and differences in quality and their influence on benchmarks

1 - Important foreword, introduction and test setup 2 - Why cyclical tests are so important 3 - Higher temperatures produce a thicker bondline thickness! 4 - Thermal expansion, viscosity, shear forces and conclusion

What you always wanted to know about thermal paste, but never got a verifiable answer: How does such a paste behave when heated and what happens when it is heated and cooled several times under the usual pressure? Completely different from what most people think! That’s why I want to show you today what pitfalls can lurk when testing paste and what can happen if some influences are not known at all. It was quite a long road to realisation for me too, but I’m happy to share the results. Because suddenly you realise that everything is much more complex than it might seem. After all, the urban legend that the layer thickness (bondline thickness or BLT for short) of a thermal paste becomes thinner the warmer the paste is and that you only have to heat it long enough to burn it in, so to speak, still persists.

I’m going to disprove this today using two very differently constructed pastes. I haven’t even picked out extreme pastes, just an expensive poser in the form of the Thermal Hero Quantum and a solid budget paste in the form of the Thermalright TF8. We will also see how cheaply mixed thermal pastes change significantly after just a few intervals of heating and cooling and where the dreaded pump-out effect really comes from.

Important foreword

My aim today is to show effects that you cannot follow in their development without equipment and long series of measurements, but you do know the consequences as the end result. This article is intended to sensitise you and help you to see what you could do differently and certainly better with simple test setups on a CPU or GPU. I have been measuring like this myself for a long time, although 7 years ago in at least 3 cycles, because I noticed certain behavioural patterns, the origin of which I could not answer at the time. But measuring a fresh paste only once must actually go wrong and I will also explain why today.

Ageing and degradation do not only occur after weeks, but from the second cycle onwards. Measurable and differently pronounced, depending on the quality of the paste. And it’s no secret that you need a few cycles before you get a usable result. The paste does not change in its primary properties through the sheer length of a test, but only through frequent heating and cooling! And this is exactly what will change the measurement results. And because not everyone can set up a test setup like this on their desk, I’ll just let you in on my work. And I can already spoil it: it’s going to be exciting!

Test setup and measurement methods

I generally measure the thermal conductive pastes for my articles and the database according to ASTM D5470-17 and I also try to reduce most negative influences in advance. This is precisely why I work with an initial layer of 500 µm after the respective calibration, which I first heat slowly to 120 °C without much pressure, then cool down to 20 °C in order to finally heat it to the constant 60 °C of my measurements. Only then do I measure the thermal resistance or thermal conductivity from 400 µm downwards in steps of 25 µm under identical laboratory conditions. This is done in a standardised way, whereby all interfering factors (such as die distortions or non-coplanar contact surfaces) can be excluded. Controlled surface conditions, unidirectional heat flow conditions, parallel contact surfaces and precisely known clamping forces are guaranteed.

I use the TIMA5 from Nanotest, a compact all-in-one desktop device that combines the measurement setup and the required PC in one device. It is a self-sufficient and, above all, automated measurement setup that I can also run in parallel to other tasks in the background. After all, who wants to sit and watch for 6 hours or more? A test series like this can hardly be realised manually. All data is saved directly to the NAS via the network. The device is recalibrated before each measurement (sensors for pressure and BLT). I normally measure the pastes at an average paste temperature of 60 degrees and a different BLT/pressure, but today I am taking different measurements between 30 and 100 °C in order to show you possible degradation in the first few cycles in diagram form.

As this all looks a bit complex to outsiders, I have placed the individual assemblies against the functional diagram so that you know where and how the measurements already explained take place. I have already explained in detail what happens in the background and how the whole thing works in the linked basic article. I don’t need to repeat all that again. Nevertheless, a little refresher won’t do any harm here either.

I will illustrate this using the already familiar diagram so that you can better visualize the meaning of the values to be determined. We can see that the effective thermal resistance affects both the material and the two contact surfaces. Yes, there are very sophisticated methods, including pulsed lasers, which can also evaluate the pure bulk value very accurately, but in practice we ALWAYS have contact surfaces. I use reference bodies with a standardized (low) roughness for the measurements so that I can also draw conclusions from these in practice. In the end, I have two values, the effective thermal conductivity and a value averaged over all measuring points of the different layer thicknesses BLT minus the extrapolated contact resistance.

External cooling is provided by a laboratory chiller from IKA, which can maintain the water temperature almost to the decimal place and which not only cools but can also reheat if necessary, so that the required 20 °C water temperature can always be strictly maintained. The hoses were connected using Festo couplings and special hoses. The room is also air-conditioned.

After so much theory, we’d better get started, so let’s turn the page and get going!