This PR updates the computation of the nuclear correction for the deuteron data set. Specifically, the procedure described in 2011.00009 has been repeated: a) a deuteron-only fit including SLACD, BCDMSD, NMCPD, NuSea and SeaQuest (and based on the n3fit methodology) has been performed and iterated (name of the deuteron fit: 210525-ern-002); b) the deuteron fit has been used to infer the deuteron correction which has been implemented in this PR. The pattern of deuteron corrections seems pretty similar to the one found in 2011.00009 w/o SeaQuest and in the context of the nnfit framework.
Here is some information relative to the deweighted case to ease this review.
deuteron to proton correction ratio (to be compared with Fig.2 of 2011.00009)
normalised diagonal elements (to be compared with Fig.3 of 2011.00009)
experimental correlation matrix (to be compared with Fig.4 left of 2011.00009)
experimental+deuteron correlation matrix (to be compared with Fig.4 right of of 2011.00009)
Here are some data/theory comparisons (obtained with 210314-001-rs-nnpdf40-1000)
Overall, all these results seem consistent to me. The nuclear shift has not changed much, while the nuclear uncertainty is reduced (because of SeaQuest and because of the n3fit methodology). Correlations are a little less strong. Because the experimental uncertainty was dominating across the board and is still dominating across the board, I don't expect any significant difference with respect to what we observed in the past. Nevertheless, I'm running a fit with these updated deuteron uncertainties before tomorrow's tag.
Note that this PR also includes the recomputation of the nuclear uncertainty for DYE605 upon addition of the missing NNLO K-factor. This is expected to be inconsequential, given that the uncertainty for the Copper target is larger than the NNLO correction.
