1. H1/ZEUS averaging meeting Sep 22 nd 2008 A M Cooper-Sarkar Studies on heavy quark scheme LHAPDF implementation

2. We need to move to using a General Mass Variable Flavour Number Scheme (GMVFN) rather than our present Zer0 Mass Variable Flavour Number Scheme (ZMVFN) You’ve seen plots from me using a GMVFN before- Robert Thorne’s scheme Labelled as RTVFN. But they were done before we chose the final specifications for our fit So let’s just clear that up first

3. Comparison of central fit plus total uncertainties to variation of heavy quark scheme: using massive variable flavour number scheme of Thorne. LEFT compared to OLDer fit RIGHT compared to HERAPDF0.1 New/Old very similar.

4. Comparison of central fit plus total uncertainties to variation of heavy quark scheme: using massive variable flavour number scheme of Thorne. LEFT compared to OLDer fit RIGHT compared to HERAPDF0.1 New/Old very similar except for very high-x D

5. Now let’s move on to ALL NEW comparisons of GMVFN to ZMVFN On the LEFT comparing HERAPDF0.1 to Thorne’s 2007 GMVFN On the RIGHT comparing HERAPDF0.1 to Thorne’s 2008 GMVFN- NOTE the difference in the gluon

6. Now compare for U D Ubar Dbar On the LEFT comparing HERAPDF0.1 to Thorne’s 2007 GMVFN On the RIGHT comparing HERAPDF0.1 to Thorne’s 2008 GMVFN- not much difference

7. Now compare for ubar dbar sbar cbar On the LEFT comparing HERAPDF0.1 to Thorne’s 2007 GMVFN On the RIGHT comparing HERAPDF0.1 to Thorne’s 2008 GMVFN- not much difference

8. Thorne’s 2007 and 2008 schemes are not that different- but the 2008 one is closer to Voica?- looks pretty consistent -apart from low-x valence- and even this isnt big Thorne says both his schemes are legitimate choices.(as is ACOT), But Thorne 2008 is closer to ACOT. 2008 also has better χ2 473..lower than my standard 476 Since all schemes are legitimate choices, perhaps we will have to live with the small difference between Thoren 2008 and ACOT as a source of model uncertainty? Voica’s plot

9. Progress on LHAPDF implementation Implementation with LHpdf method from parameters by evolution ‘on the fly’ u-valence and d-valence Experimental errors only Compare to our own plots- perfect agreement

10. Implementation with LHpdf method from parameters by evolution ‘on the fly’ u-sea and d-sea Experimental errors only Compare to our own plots- perfect agreement

11. Implementation with LHpdf method from parameters by evolution ‘on the fly’ gluon Experimental errors only Compare to our own plots- perfect agreement

12. Model errors will be done with the LHgrid method Currently checking out agreement of grids with LHpdf method And checking grid interpolation methods Blue line is (LHgrid-LHpdf)/LHpdf for linear interpolation Green line is the same for 4 th order polynomial interpolation Axes are in % - so differences are much less than 0.2% !! Quadratic is better than linear interpolation- important at high-x Red line is linear interpolation but only in Q2 (uses grid points for x) so my grid spacing at high-x is the problem for linear interpolation Agreement gets better as Q2 increases and goes up as Q2 decreases, but is still within 0.2%. We are nearly there!

13. Look just at high-x 0.1 to 1.0 for Q2=10 ans Q2=10000 Linear is only good up to ~0.4 quartic is good up to 0.8-0.9