search

NNPDF / pineappl

PineAPPL is not an extension of APPLgrid
https://nnpdf.github.io/pineappl/
GNU General Public License v3.0
12 stars 3 forks source link

Add fastNLO importer #25

Closed cschwan closed 2 years ago

cschwan commented 4 years ago

Similar to issue #17.

enocera commented 3 years ago

@cschwan It would be nice if we could make the fastNLO importer also work with DIS grids. One specific example is DIS+jet(s). Apparently, the grids available from Ploughshare have been removed (for instance, they were available at https://ploughshare.web.cern.ch/ploughshare/record.php?group=applfast&dataset=applfast-h1-incjets-fnlo-arxiv-1406.4709, but are no longer there). I suspect that the reason is that a bug in NNLOjet was discovered. Anyways, for the sake of looking at the fastNLO converter, I think that we could use the grids that I've got from Daniel Britzger some time ago. I've (temporarily) put them at this link https://data.nnpdf.science/fnlo_tables.tgz

enocera commented 3 years ago

The link includes several experiments from H1 and ZEUS, if the importer works with one grid, it should with all of them.

cschwan commented 3 years ago

Great, thanks a lot. I'll have a look at it soon.

cschwan commented 3 years ago

I've started working on this, and commit ea81006f7f936fb70ebcb5dde6793afc46ce57bf adds a program fnlo2pine that can already read fastNLOCoeffAddFix tables. The numbers are still wrong, but not too far off, so there should only be a few bugs left to find. I'm also confident that reading out fastNLOCoeffAddFlex-type table should work similarly - this is the format of the tables you've linked above.

cschwan commented 3 years ago

Starting with commit 6a0b4f5be494d5ba59f4fabe1e117778d55a0042 we can now convert one-dimensional fastNLOCoeffAddFix-type tables, for example Mitov's ttbar tables - all of them are working.

cschwan commented 3 years ago

Commits 0e18c2235c68cf3366dfb62b8c9670ba97e2e3cd and 8a5c7b947a251c0e2bcfcba86162c100b565d906 add support for symmetrised fastNLO tables, such as the 2D and 3D jet cross sections that we're using.

cschwan commented 3 years ago

What's left to do for this issue is to implement support for the flexible-scale grids (these: https://data.nnpdf.science/fnlo_tables.tgz), which support different values for the factorisation and renormalisation scales. If I recall it correctly the linked DIS grids set the factorisation scale to Q^2 and the renormalisation scale to sqrt(Q^2 + ptjet^2). At the moment this unfortunately isn't supported in PineAPPL (nor is it in APPLgrid).

cschwan commented 3 years ago

Preliminary support for flexible-scale grids (the flexibility to change the scale function isn't retained, but independent renormalization and factorization scale choices are supported) has been added in commit c71dd92012250f61e96e9e30253fa11cf4b7e845. I tested the following sets of tables from ploughshare with fastNLO 2.5.0:

All of the grids are successfully converted for the central scale choice, however, scale variations need a bit more work.

cschwan commented 2 years ago

Commit 5569124b243f7d600d296aa9ac7779091952e6e4 introduced a regression, which causes the conversion of fastNLO tables for dijets a single-jet inclusive to fail (numbers do no longer agree).

cschwan commented 2 years ago

The previously mentioned problem is fixed in commit 628472a4ce8c6b9a000456200ce481faac21bb3c.

cschwan commented 2 years ago

The remaining development is be found in #120 and subsequent Issues/PRs.