A few years back I built some thin film superconducting transformers
using Nb spiral inductors separated by PECVD dielectric.  For the
dielectric I first tried a PECVD "oxide" using SiH4 and NO precursors.
The resulting film had terrible step coverage with dogbones and
electrical shorts where the dielectric crossed the Nb lines.  I then
tried an oxynitride using SiH4 and N20 which had much better step
coverage and produced good transformers.

Have you looked at the step coverage with a SEM?  Your problem may be
simply pin holes in planar areas or it could be a step coverage problem.
Using TEOS rather than SiH4 will likely give better step coverage.

Roger Shile

-----Original Message-----

  We are fabricating a microcoil using two metal layers and PECVD oxide
for electrical insulation. The metal layers are gold deposited using
e-beam evaporation with a few nm of Cromium for adhesion. We found that
the PECVD oxide that grows on top of gold is VERY ROUGH (tipically a
roughness of 50nm for a 200nm thick layer). Further it is plenty of
pinholes. Does anyone know the reason?

We use N2O and 5%SiH4 95%He as precursor gases, deposition temperature
is T=300ºC