[volt-nuts] low emf solder

[Florian Teply]

Am Sat, 29 Oct 2016 20:41:48 +0200
schrieb Attila Kinali <attila at kinali.ch>:

> On Sat, 29 Oct 2016 15:27:49 +0100
> "Dr. David Kirkby (Kirkby Microwave Ltd)"
> <drkirkby at kirkbymicrowave.co.uk> wrote:
> 
> > But if one could use copper as the bond wire, rather than gold
> > which is quite common, then it would give you very little thermal
> > EMF  
> 
> Copper is more and more used for bonding wires.
> See:
> http://www.ti.com/lit/wp/sszy003/sszy003.pdf
> http://link.springer.com/article/10.1007/BF03214983 
> https://nepp.nasa.gov/files/26611/2015-370-Rutkowski-Final-Paper-NEPPweb-Copper-Wire-Bonds-TN26444.pdf
> 
> But be aware, that the bond pad on the chip is still aluminium in
> most cases I don't know whether there is any process that uses copper
> bond pads at all. As far as I know, all of the relevant processes for
> analog designs (ie those with node sizes over 130nm) still use
> aluminium interconnect only. IIRC the first node size that got copper
> interconnect was 65nm and about 10 years ago.

I seem to recall that either IBM or AMD used copper already in 90nm,
but that's still in the same ballpark. And I don't see a reason why
one shouldn't use copper bondpads if the better part of the
metallization is copper already to begin with. But on the other hand
with the aluminium backend you mention for 130nm and larger node sizes,
you're also in for some interesting material combinations: a) the
aluminium isn't pure but usually has a copper content of between 0.5
and 1.5%. Then b) the vias are usually not aluminium but Tungsten or
something similar. Aluminium is only found at nodes starting from about
0.5 micron upwards, with 3 or less metallization layers. And finally,
as aluminium doesn't stick too well to Silicon dioxide, there will be
thin layers of, say, titanium or titanium nitride in between insulator
and metal. 
And please don't underestimate thermal gradients on chip. It's not
uncommon to see several tens of degrees difference in surface
temperature across even a small chip...

Best regards,
Florian