martin-ueding
commented
4 years ago This cutoff is implemented exactly like in the contraction code now. We have the following choice:
ContractionMomentumCutoff[0] = 4;
ContractionMomentumCutoff[1] = 5;
ContractionMomentumCutoff[2] = 6;
ContractionMomentumCutoff[3] = 7;
ContractionMomentumCutoff[4] = 4;Then we just filter the individual momenta by this. If some momentum occurs that is above the cutoff we just discard that combination and therefore row/column from the correlator matrix. The actual filtering is done with some infix heavy code, which one could make even worse :wink:
Select[
...,
MomentaMaxNorm @ # <= ContractionMomentumCutoff[Norm[totalMomentum]^2] &];
kostrzewa
maowerner
While trying to reproduce the rho projections I encounter a bunch of failures due to missing HDF5 datasets. The first one missing for the various cases are the following:
The
000in theT1uis a different case. There Markus has implemented an individual momentum cutoff atp_i^2 ≤ 2whereas I do not have this limitation and therefore it tries to access a higher momentum. Perhaps I add this rule into my analytic projection code at some point.The reference momenta (000, 001, 110, 111) work just fine, but these other are problematic. I would guess that I screwed up the reference rotation.
In the
-110in theA2has this as the first missing one:The momenta on source and sink are these (reverse the ones at sink):
Momentum conservation holds. But notice how the fourth momentum has norm squared 6, so it certainly isn't computed. Eyeing the table one can see that all the problematic ones have norm squared 5 or 6. For the
T1uit is problematic at norm squared 3 already, but we know why.So there is a need to implement the individual momentum restrictions into the analytic projection code as well in order to prevent these data sets from being requested. I am a bit surprised that the reference momenta work, though.