Test equipment for material tests on thermally conductive pastes: microscopy and laser-induced breakdown spectroscopy (LIBS)

1 - Thest methods and equipment 2 - Microscopy and material analysis

Microscopy: Tear-off test

If you spread a paste with a spatula on a smooth surface (glass slide) and then pull it off thinly until the layer tears off, you can already conclude two things. You can see how well the paste adheres to a smooth surface such as a GPU die and you can see how it tears off at the resulting edges. Either this is a smooth transition and you can even see different grains that are pulled further, or it peels right off. In extreme cases, the silicone is even pulled out of the paste. This paste from the first example is relatively “grainy”, but can still be spread reasonably well. The compromise between handling and performance is quite good and the paste is not crushed into its components.

However, you can already see from the islands that the paste does not flow very well and you should at least pay close attention to a homogeneous surface without cracks when applying it. But this is a general characteristic of highly filled pastes. You just have to know it in order to act accordingly. What you can already recognize at this low magnification are the sometimes quite large particles.

The exact opposite is a rather poor paste and no, this is not raspberry ice cream, but a colorful thermal paste. Even in the small resolution we can see the rather cream-like consistency of the paste. It’s easy to spread and of course gives beginners a nice feeling of security, but that’s about it.

Under magnification, you can already see the shiny binder. Yes, it is a real beginner’s paste, but it is not durable. You won’t be happy with a paste like this for long. Unfortunately.

Microscopy: Particle sizes

This is where I reach the limits of the resolution of the Keyence VHX-7100, because in the sub-micron range and with nanoparticles I can see that there is still something there, but even at 2000x resolution it is no longer possible to identify or even measure such small particles cleanly. I even go back to 1500x magnification to rule out possible image errors (shadows etc.), which the image stabilization may then sell to me as grain. However, the somewhat coarser grinding degrees of some particles can be easily analyzed and determined. In the paste from the first example, these are also due to the performance and you can definitely do this if there are enough smaller particles for the gaps. And they are.

Of course, it is also a question of cost which grinding grades a manufacturer uses and, above all, which tolerance limit is applied. A really uniform grain size can cost 20 times as much as what you can buy cheaply as a mixture of different sizes. Which brings us back to the shelf life of such a paste. The siloxanes used also vary in quality and price and it is certainly the case that if you fill a matrix with thermally conductive particles, these can only be incorporated very differently depending on their size. If, for example, you optimize a matrix to the range of one micrometer, any significantly larger (or smaller) particle will have a very negative effect on the structure and consistency. The particles shown here indicate a medium-performance and also only medium-durable paste, which, however, still does what is expected.

The grinding grades of the particles from the colored paste, on the other hand, are very fine, but there is far too much silicone in between. One could assume that nanoparticles of ZnO, which are no longer visible to the microscope, have been slurried into the matrix instead, but are they really there? I can also answer that later. In any case, it would have made sense in terms of performance to simply fill the paste a little higher instead. Well, that would have cost more money and probably also more time for the mixing process, which ultimately amounts to the same thing.

Material analysis

The first paste relies on significantly more aluminum than many other pastes, so that the ratio to zinc is approximately 5:1. But I also need to say something first to help you better understand the weights determined, because LIBS is used to determine the weights of individual chemical elements and not compounds. The aluminum found is a component of the contained _Al2O3and thus binds a large part of the listed oxygen, just like the zinc oxide ZnO. Then there is the matrix, in this case siloxanes, where you find molecular chains of silicon, again oxygen and also hydrogen. Only the approximate amount that remains according to the calculator would really be an inclusion of suspended air and, for example, water that was not outgassed during the manufacturing process in the vacuum chamber. In the case of the paste in the example, it can be assumed that there are no air or water inclusions to any significant extent. It is therefore also a kind of quality test.

With the second paste, however, we can see how NOT to do it. The supplier should have a word with the OEM, because with well under 40% metal content and over 60% residue (over 20% silicone oil plus the bound oxygen in the oxides), it is very clear that the paste was only filled to a very, very low level, which I had already suspected in the microscopy. And it also disproves a possible filling with many nanoparticles, because then the zinc content would be significantly higher. It is slightly higher than in the first paste, but not high enough to even begin to ensure performance. And the whole thing is not durable either, because the paste will simply bleed out as a soup in this consistency. You don’t even have to be an oracle.

Summary and conclusion

In direct interaction with my measurement according to ASTM D5470-17, you can actually estimate every paste pretty accurately. More would actually only be possible in the laboratories of the really big manufacturers, but I am unlikely to put on this shoe, of course also for financial reasons. After all, the purchases of the last 12 months alone are equivalent to the value of a terraced house and at some point, common sense and reason set hard limits. But if you do something, and I learned this from the disaster with the botched fan test by a third party, then you really should do it yourself and, above all, do it properly. Always from the point of view of your own standards, of course. I learned a lot from that, including financially.

But I’m already planning for the future (including that of my children as possible successors) and am counting on the fact that thermally conductive materials (TIM) are becoming increasingly important in power electronics applications. They are all long-lasting laboratory devices, combined with maintenance contracts and long warranty periods through to the guaranteed availability of spare parts, with which you can do many things, not just testing thermal pastes or pads. That’s a spoiler for things to come, which should be no less exciting. And I’m pretty sure that there will still be a certain need for information about such in-depth things in the future, even outside the TikTok bubble. But in order to fill such a gap, you have to offer something reliable to stand out from the crowd.

Of course, you can moan about the current seemingly irreversible shift in focus from the elaborately written word to shallow entertainment short films, or you can look for and fill a niche that such media can and will never fill. That’s exactly the plan and so far I can’t report anything negative. When I look at the trends in the community and the age structure, it’s not just old, white men who are actively involved, which gives me hope. Just a personal comment and conclusion to today’s article.

Read more: