Diamond Thermal Conductivity: A New Era in Chip Cooling

(spectrum.ieee.org)

64 points | by rbanffy 5 days ago ago

29 comments

  • lateforwork 4 hours ago

    > But my research group at Stanford University has managed what seemed impossible.

    Wait a minute, others have been doing this already: https://www.df.com/

    How is this different?

    See also: https://youtu.be/ggQKZDZsDec

    • mutagenesis 35 minutes ago

      What you linked is bonding a diamond substrate to the back of your IC. What's in the post is growing diamond lattice/features directly on your wafer. With the new way, you can get diamond closer to your heat sources, increase contact area, etc.

      No idea if it actually matters. Is this a single digit percentage increase in thermal conductivity by messing with a finicky, temperamental process? I don't know. What the paper writers are proposing is under the limit of when transistor structures break down, but not by much.

  • 0cf8612b2e1e 5 days ago

    Have diamonds found their way into other industrial cooling solutions? With the research into gem grade diamonds, I have been expecting cheap ugly synthetic diamonds to be used in more products. I have long joked that I want a diamond frying pan.

    • gpm 6 hours ago

      3D printer nozzles, which is sort of the opposite (industrial heating products).

      Part of the argument is that better heat conduction means that you can run the nozzle cooler resulting in less heat conduction to the cold side (above where you want the filament to melt) so I guess its "cooling" in a sense too.

    • Atomic_Torrfisk 10 hours ago

      > ugly synthetic diamonds

      Not any more, their quality has increased recently. Not that I care, wife and I did without them during our engagement.

      • jrk 7 hours ago

        I think the point was not that gem-grade synthetic diamonds are ugly, but that, as industry masters gem-grade production, presumably below-gem-grade production (“ugly synthetic diamonds”) would become cheap enough to deploy in more engineering settings where diamond’s other unique properties were the key concern.

    • Yossarrian22 6 hours ago

      I’m surprised nobody has done a phone screen yet

      • Tuna-Fish 4 hours ago

        Large single crystal diamond, what is required for a nice transparent screen, is still quite expensive. This article is about polycrystalline diamond, which is not really that transparent, but is nearly as good at thermal conduction as monocrystalline diamond.

      • rowanG077 2 hours ago

        Even if that were cheap I don't think diamond would excel in this use case. It's of course extremely hard but I'd expect that it would be extremely prone to cracking. In addition the high index of refraction would make the diamond screen very reflective and you would need some fancy coating which of course wouldn't be as strong as diamond.

    • rbanffy 3 days ago

      > I have long joked that I want a diamond frying pan

      As long as you don’t use it on a gas stove, you should be fine.

      • kees99 10 hours ago

        Why? Diamond has very low thermal expansion, so no risk of stress/embrittlement/cracks from uneven heating.

        Or you mean it'll catch fire? Also not a concern. That is supposed to happen at a temperature well above anything useful for cooking.

        • rbanffy 9 hours ago

          The temperature of the blue flame on a stove should be above 1000 Celsius, well above what’s required to oxidise diamonds. They won’t catch fire, but your diamond pan will erode. Once you remove it from the flame, it won’t continue “burning”.

          Should be safe on electrical stoves though.

          • kees99 8 hours ago

            Would that >1000°C reach the surface though?

            There are some heady boundary-layer effects and temperature/temp-conductivity gradient physics involved here. For simplicity sake, consider a plastic [1] bag full to the brim with water, held over open flame. Will bag melt (oxidize, erode)?

            [1] polyethylene melts around 120-ish °C and ignites around 220-350 °C (sources vary)

        • LtdJorge 9 hours ago

          I think because it will burn your hand?

          • rbanffy 9 hours ago

            It’s safe to assume the handle is made from something else.

      • PunchyHamster 5 hours ago

        Just put them on inside only

  • Klaster_1 7 hours ago
  • snalty 9 hours ago

    This reminded me of this: https://www.innovationcooling.com/products/ic-diamond/?srslt... which seemed to be all the rage in PC building 10 years ago

  • lightedman 6 hours ago

    Nothing new, Applied Diamond has made this stuff for several years and it is incredible. Imagine putting a 15w LED on a typical 20mm star board made of diamond - you do not need a heat sink. Just minor air flow over the package is enough.

    A little unlike IEEE to be nearly half a decade out of the loop.

    • akshatjiwan an hour ago

      I think what they have grown diamond on the transistor which then bonds to the substrate through a SIC interlayer.

      From what I understand their idea seems to be that since most heating occurs at channels they act like hotspots and therefore it would be much better to drain away heat from them directly.

      This is different from creating transistors on a diamond substrate.

    • Tuna-Fish 4 hours ago

      The invention is a fast, low-temperature deposition process, which can then be used directly on a semiconductor device.

      • lateforwork 3 hours ago

        Diamond Foundry achieves the same end goal even if the deposition methods and technical processes may differ: https://df.com

        • Tuna-Fish 2 hours ago

          They produce the diamond wafers separately and then later bond them, a process that is dramatically more expensive and results in a worse bond.

    • yorwba 5 hours ago

      The new thing here is growing a thin layer of diamond directly on top of a chip.

      • trhway an hour ago

        naturally, graphite has similarly high thermal conductivity along the layer direction (which is basically graphene), and one would think that there should be some way to put such a thin layer of graphite/graphene on top (or inside) the chip to achieve similar results.