I thought I would throw this into the mix - as a sequel rather than postscript to a rather silly recent thread on this matter. A few years ago, it was impossible or impractical to find a suitable thermowell for use with 20mm copper pipe, so I decided to make them. They are simply a brass plug in a standard T fitting, with a flat-bottom hole for the sensor and sealed with heat-sink paste. I quickly became unsure about just how bright an idea this is, and now I am even less sure. I raise this particularly as other forummers have advocated using heatsink paste, Arctic Cooler and the like.
I don't think Arctic Cooler heatsink paste is fit for purpose. This is the stuff the game boys used under humungous copper heatsinks clamped down onto their CPUs. I reckon the operative word is clamped, and I am using it as filler. There is less than 1mm of slop in my installation, but it is still filler. Further, while everything looked fine when installed and it hasn't been touched, it certainly doesn't now. It looks like the paste has evaporated.
I have only one thermowell in service, as I quickly concluded that having the sensor simply held against the pipe in a dab of Arctic Cooler and secured under an insulation sleeve works just as well. It's just that a thermowell looks more (ahem!) professional, and I wanted to impress the powers that be.
I now find that this installation is also a long way from something you would want to show your mother. The Arctic Cooler is dried up and looks powdery.
The real objective is to keep air out of the game, and I therefore submit that silicon goo is a better solution. It is specifically not intended to be used in a thin smear, nor to be used between clamped surfaces - indeed quite the opposite, i.e. kosher to be used as a filler.
I haven't found anything to suggest that a thermowell is a better way to go than simply lodging the DS18B20 against the pipe under an insulating sleeve. The problem of excluding air applies to both. The dregs of the paste is easily removed with metho.