A lie or a gray area? How to construct high thermal conductivity and why the information provided by sellers is almost never correct

We have discussed this many times before: The thermal conductivity figures for thermal pastes on packaging and data sheets are almost always far too high. Of course, this is also due to the fact that either unsuitable measurement methods are used or the conditions are changed in such a way that the values determined are far too high. This is exactly what I would like to take up again today, because the importers of sachets in particular blindly believe what the Chinese OEM is filling them with for a pittance. High values sell well, of course. But at the end of the day, the customer is the one who is stupid, because he can’t measure it. Today we will take a look at a paste with a legendary 17 W/m-K.

Of course, I can’t check whether anything was measured at all with this paste or whether the OEM or bag sender just agreed on a printed value for marketing purposes. But we can think about how you get such a value without firing up the dragster. Quite legally and as the result of a measurement that delivers a misleading result from the customer’s point of view, but where one could not even assume intent. Yes, there is such a thing. And far too often, because this is where the bucket of paste I keep quoting comes into play.

The disadvantages of the ASTM Hotwire measurement

The ASTM Hotwire test for measuring the thermal conductivity of thermal pastes, which is actually quite inexpensive to implement, has serious disadvantages that can quickly affect the accuracy of the results. One of the main problems with this method is the potential misestimation of the actual thermal conductivity of the materials tested, which often leads to values that are far too high. One of the main disadvantages of the ASTM Hotwire test is the sensitivity of the test to the contact resistances between the thermal paste and the metallic plates. These contact resistances are caused by unevenness on the surfaces of the plates and the imperfect distribution of the paste. Even small air bubbles or uneven layers can impair the heat flow and thus falsify the measurement results. As the test is based on the assumption that the paste conducts the heat flow evenly, such irregularities can quickly lead to systematic errors and usually do so for materials above 1 W/m-K.

In addition, insufficient consideration is given to the influence of temperature gradients along the metallic plates. If the temperature distribution is not uniform, this influences the temperature measurements and leads to an inaccurate calculation of the thermal conductivity. In addition, the hotwire test can be further distorted at high temperatures due to the thermal expansion of the materials, which can also lead to incorrect results.

The often too high measured values also result from the so-called “guarded hot plate” design of the test, in which the heating of the wire can lead to an overestimation of the thermal conductivity. In practice, this means that the test overestimates the thermal conductivity of the thermal paste because the method is sensitive to local overheating of the wire. The wire generates a localized heat source that is not necessarily representative of the uniform heat transfer in real applications. The shortcomings are therefore due to a combination of contact resistance, uneven temperature distribution, and the specific test set-up, which do not always correctly replicate the real-life operating conditions of the thermal pastes. Therefore, these measurements are less reliable for more conductive pastes and can significantly overestimate the actual performance of the thermal conductive materials. But they are inexpensive, as such a system ultimately costs only a tenth of the price of a proper tester.

How to do it right

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 to 20 °C to finally heat it to the constant 60 °C of my measurements. Only then do I measure the thermal resistance or thermal conductivity under identical laboratory conditions at a constant 60 °C average paste temperature and from 400 µm downwards in steps of 25 µm.

This is done in a standardized 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. It is therefore anything but the simple bucket method of the usual suspects.

Correct measurement vs. simulation of an incorrect measurement

I am now also trying to measure the paste “incorrectly” by creating inappropriate conditions (temperatures, extremely high pressure and therefore a tiny Rth) and calculating this so-called bulk thermal conductivity using only a few measuring points as in the hotwire test. I also leave out the interfering points here. This comes relatively close to a “bucket test” in terms of inaccuracy. We can see the protocol of my degradation test of this paste from the packaging in the introduction from above and I also arrive at just under 17 W/m-K. But only because that’s what I want. I could easily print the result on the packaging and be delighted, but this value is completely unrealistic:

If, on the other hand, I now measure with a total of 6 temperature sensors and one for the heater, then carry out a total of 17 individual measurements with different values for the bondline thickness (BLT) (and the two test bodies on the heater and cooler have also been measured and calibrated), then significantly more accurate values are obtained. This is because each individual measurement has a threshold for the temperature windows of all 7 sensors, the BLT and the pressure over 100 samples of measurement data and must have maintained this window for ALL values for at least 2 seconds.

The effective thermal resistance generally changes linearly to the BLT, so I remove some values from the calculation that are not exactly on the imaginary line. Then, logically, I also get the correct, adjusted value for the remaining 13 measurements, which may even be slightly higher:

So this paste (Thermal Hero Quantum) has a real bulk thermal conductivity of 3.16 W/m-K and not 17 W/m-K! And now let’s compare something, as malicious as we are… Dow Chemical specifies at least 5 W/m-K for the DOWSIL TC-5550 and writes internally of up to 5.3 W/m-K. According to my method, I measure around 5.28, which should certainly be sufficiently accurate. If I were to misjudge this paste now, I would even be at an incredible 20 to 25 W/m-K. For anyone who would like to see this again in curves, here is a comparison of the thermal resistances of the Thermal Hero Quantum and the DOWSIL TC-5550:

And once again for the effective thermal conductivity taking into account the existing interface resistors:

I think that’s enough theory for one Friday, but I really wanted to get this off my chest.