A new technique for putting a metal-layer heat-sink thermal interface directly on the silicon die aims to eliminate the need for thermal grease or pads.
Heat sinks are an unavoidable fact of thermal life and design for many ICs and discrete power devices. Without these often passive and sometimes active devices, self-generated heat would not dissipate to that magical kingdom of “away,” where the excess heat would instead be a problem for someone or something else, and devices would cook themselves to early death.
But attaching the heat sink to the device to be cooled is often not simple. Yes, there are many commercially available heat sinks (Figures 1, 2, and 3 are just a few of the nearly infinite choices) with many tailored to specific device packages, and even customized designs which maximize convection, but that’s only part of the solution. It’s critical to keep the thermal impedance between the heat source and its sink as low as possible. Any voids or imperfection in the thermal path will reduce heat flow and thus heat sink effectiveness.
This is an old problem, of course with two often-used solutions; use thermal grease or a thermally conductive pad (see the Henkel/Bergquist site, for example) between the package and the heat sink. Either solution adds cost and assembly effort; if you have ever applied it, you know that thermal grease is tricky to handle and apply smoothly and evenly. Regardless of the chosen approach, you still need the heat sink, of course (The reference, at the end of this article, is a good overview of thermal interface materials and a highly accurate fixture developed for testing their properties).
Some ICs have a thermal pad under their package to allow using the PC board itself as a heat sink. That’s also an effective technique for some package sizes and thermal loads but adds cost to the IC package and often assumes the PC board is not also trying to cool many adjacent devices; further, that bottom-side thermal path sink may not be sufficient for a hotter IC.
Recently, a research team at the Binghamton University (NY) Mechanical Engineering Department devised and implemented a different approach to enabling heat sinking. They used selective laser melting (SLM) and additive manufacturing to tightly bond a tin-silver-titanium alloy (Sn3Ag4Ti) to the silicon. This formed a thin titanium-silicide bonding layer that acted as a glue between the silicon chip and the metal alloy. Using this technique, various heat-removal devices such as vapor chamber evaporators, heat pipes, and micro-channels were directly “printed” onto the electronic package without using any thermal interface material.
Printing the microchannels onto the chip was not a straightforward task. Most metals and alloys will not form a good bond with the silicon due to poor adhesion with silicon and thermal expansion mismatch. As explained by Assistant Professor Scott Schiffres, leader of the project team, “We print microchannels on the chip itself to make spirals or mazes that the coolant can travel through directly on the chip instead of using the thermal paste as the connection between the heat sink and the chip.”
Implementing this direct additive printing is not easy, and the bonding between the metal alloys generally used in additive manufacturing and silicon is relatively weak and has high contact angles (poor wetting and interfacial strength). With this interlayer material and application technique, the wettability and reactivity with the silicon substrate was increased significantly.
Unlike use of brazing, which can take tens of minutes to form a strong bond, they were able to establish good bond in sub-milliseconds (Figure 4). They did this by using intense laser heating via an alloy that can form a strong intermetallic bond to the substrate at a low temperature, while exposing the sample multiple times to give sufficient diffusion time for a strong bond. The bonding to silicon is due to the formation of a micrometer-thin titanium-silicide interfacial layer that makes the silicon wettable to the Sn3Ag4Ti layer, and the relatively low melting temperature of Sn3Ag4Ti (about 250°C) reduced thermal stress on the die.
Their test results were positive and significant: they claimed it yielded a 10°C improvement in die temperature compared to a heat sink using conventional thermal interface materials. They also measured other properties of the material including the Sn3Ag4Ti bulk thermal conductivity, which required a complex setup (Figure 5).
For more details on the entire process and results, you can read the full technical paper “Additive laser metal deposition onto silicon” published in Elsevier’s Additive Manufacturing. The less-technical news article by Binghamton University (and video below) devote considerable space to quantifying – to several significant figures! – the potential positive impact on climate change and carbon dioxide emissions of this approach (which seems like quite a statistical reach to me).
I have no idea if this innovation will be commercially viable. Perhaps it will be like the many “breakthroughs” in battery chemistries and electrode technologies that looked great in the lab but did not scale to pilot or mass production for various reasons. Or perhaps it will become viable, but only after many more years of refinement (often cited as a sudden success that was years in the making).
What’s your sense of the practicality and desirability of this approach to making the heat sink interface an integral part of the die?
Bill Schweber is an EE who has written three textbooks, hundreds of technical articles, opinion columns, and product features.