Hi Boud,
I've just had such a though.
If we assumed some multiply-connected space-time within some model (eg. dodecahedral) then my question is whether is leads to nongaussian (NG) features in CMB. I think it definitely should. Such spacetime, gives a modified power spectrum of fluctuations, limited to fundamental domain size and there should be repercusions of this size in harmonic modes that are natural multiple of that size. Then given a feature at one length scale should automatically give features at other scales - hence vialation of gaussianity. This could be investigated by bi- or tri-spectrum optimalized for these particular perturbation modes that come from the model. Given NG simulations one could estimate the sensityvity of this approach. This is also a test of the multiply connected space hypothesis.
Is that right ?
What I would need is to have a full fourier space of perturbations with topology encoded in it.
pozdr. Bartek