search

NNPDF / pinecards

Runcards needed to generate PineAPPL grids for NNPDF processes
3 stars 1 forks source link

Update for VRAP 1.4 #164

Closed scarlehoff closed 1 year ago

scarlehoff commented 1 year ago

This is just a bump up for vrap 1.4

I've added a new line to the theory card so the same one can be used with pineko, can revert it if needed.

As a draft for the time being since I'm looking into the E906.

scarlehoff commented 1 year ago

This version agrees 100% perfectly with theory 200. Current theory 400 had a problem in the last bin (and was a bit bumpy bin-by-bin).

The check below is validphys theory 200 (left) against the grids (not the fktables) generated by this branch with vrap 1.4 (so with @cschwan fixes) at NLO. The PDF is NNPDF40_nnlo_as_01180.

       200         vrap         ratio
0  1.151581  1.153289  0.998519
1  1.172257  1.173061  0.999314
2  1.162259  1.162165  1.000081
3  1.121157  1.120979  1.000159
4  1.054628  1.054089  1.000512
5  1.009504  1.010595  0.998920

For completeness, same comparison code with NNPDF31:

0  3.042647  2.599038  1.170682
1  2.044257  1.933439  1.057317
2  1.886642  1.816957  1.038353
3  1.948976  1.855019  1.050650
4  3.421160  2.738610  1.249232
5 -0.492614 -0.645222  0.763479

My guess (/cc @cschwan) is that the fixes to vrap have a sizeable effect here because it is a composition of 20 fktables. In the same vein, instabilities have a factor 20 as well :P

Of course, the NNLO correction I find with vrap is not compatible with the cfactors computed with NNPDF31 but I also don't trust those.