Hello,

I can just recommend (have not tried by myself) to solder using low melting point solder especially dedicated for low temperature use. See Wikipedia for solders available. As for test, even Rose alloy may work well. And it can be cheap test material. Melting point ~100 C should be low enough to prevent outer layer catastrophic burn out. V-groove is absolutely necessary I think.

Electronics standard SnBi alloy could be used if fiber withstands temperature up to 150 C.

Some laser companies uses low temperature solders to fix adjustable mounts in lasers after final alignment. Heats-melts if necessary and readjust. Say "no problems", but some are, therefore is not so popular.

There are some other dangers. For example when InGaAs will be glued on Cu they will form "bi-metal" plate which will bend when cooling.

Use a scotch tape. It will work.

V-grooves and several drops of soft glue, proved

Thank you all for replying.  I hope to test the V-groove and glue technique soon, hopefully then this will give more stable behavior.  Even the glue by itself is fairly stable when tested by cooling in LN2 .

I hadn't considered solder, would there be an issue with this cracking as it cools however?

I didn't mention the temperature specifically but this must be stable down to 4K.  The fibre is also single mode with a core diameter of ~5um so stability really is critical.

Have no experience with such low temperatures. However, someone did part of job for us:

http://archive.org/stream/nasa_techdoc_19630013160/19630013160_djvu.txt

"DISCUSSION OF PROBLEMS AND SPECIAL TECHNIQUES

In the design and construction of scientific equipment of this

type, many unforeseen problems arise which must be surmounted prior to

obtaining the desired data. Initially, much difficulty was encountered

in attaching the vacuum chamber to the heat sink. Attempts were first

made to seal the chamber with Rose's alloy. In using this alloy, some

sealing operations would hold a vacuum at room temperature; but, when

subjected to liquid nitrogen temperature, the seal would rupture. The

ruptured seal would then close on returning to room temperature. Cerro-

bend was tried with only slightly better results until the technique of

preparing the two mating surfaces was changed. The knife edge on the

heat sink and trough on the chamber were cleaned with Green Streak Flux

and tinned with 50-50 lead-tin solder. Molten cerrobend was then

poured into the preheated trough. The tinned knife edge was also pre-

treated, and the flux and oxides wiped from the surface before it was

immersed in the cerrobend filled trough. A stainless steel tool was

worked around in the molten cerrobend to remove any trapped air or

oxides that had formed. On cooling, this technique produced a high

vacuum seal that would withstand thermal cycling from 20° to -196°C

(68° to -321°F)."

It can be expected that Rose or others similar composition are of moderate applicability. If nitrogen can be isolated, as shown in work above, so must be no cracks or micro structuring... I expect. However there is significant difference of materials because Rose is non-eutectic and Cerrobend (Wood) is eutectic. Microstructuring may be less expressed in eutectic alloys. But these speculations are just basics of alloys not tested in practice. Maybe someone can tell better and for real.

Oliver, hard to predict the fiber behaviour at 4K, but I think soldering  will not work because of very different thermal coefficients with fiber (metal). For  5 mkm you anyway have to provide an optical system with fine adjustment. Good luck!

Agree with Alexander Apolonskiy.

V-groove + glue -- tested & proved

Two skilled persons claim gluing is the best. Go for it. Just get the answer what kind of glue should be used. Except you will address this question to different auditorium - low temperature specialists. They may have something to add.

I would first propose without glue, only with a sharp groove

I'm not so sure about your geometries, but my first thought is that fixating the fiber to the copper is maybe not the best way. If your alignment (fiber to chip) is critical, you need to take into account that glass and GaAs have different thermal expansion coefficients. So if both are stuck to the copper, the points where they should couple might move with respect to eachother.