H1/ZEUS fitters meeting Jan 15 th 2010 Am Cooper-Sarkar Mostly about fitting the combined F2c data New work on an FFN fit PLUS Comparing HERAPDF to Tevatron data for fun Z-rapidity and W lepton asymmetry

HERA combined charm data compared to HERAPDF with various mc values. The variation mc=1.6, Q 2 0 =2.5 describes the data quite well (recall Q 2 values < 2.5 are not fitted).

So what happens if we actually fit the charm data? First with the RT-VFN formalism NOT Fixed flavour Number With scales all Q2 as appropriate to that scheme Use the usual cuts i.e. only data Q2 > 3.5, so 41 F2c data points Assume the two correlated errors supplied by Katerina are NOT correlated to Any of the 113 errors of the HERAPDF1.0 fit Assume normalisation of charm data wrt HERAPDF1.0 is 1.0 NO error Three fits: mc=1.4 as for HERAPDF1.0 central fit mc=1.65 as preferred by charm group, HERAPDF1.0 upper limit mc free – one of the fit parameters

mc=1.4GeV χ2/ndp = 731/633 for HERA-I combined 596/592 for F2c 135/41 This is a DREADFUL fit to the charm data And the HERA-1 data are also worse fit than for HERAPDF1.0 (574/592) So try mc=1.65GeV χ2/ndp =627.6/633 for HERA_I combined 584/592 for F2c 43/41 Clearly a more acceptable fit for both data sets Finally let mc be a free parameter mc=1.54 ± 0.09 χ2/ndp=627.3/633 for HERA_1 combined 579/592 for F2c 48/41 A compromise which fits HERA-1 data better and F2c slightly worse The slight tensions between F2c and HERA-I inclusive may of course be due to the use of FFN in the f2c extraction, and RTVFN in the fit—

The summary plot illustrates the two fits which include F2c data using mc=1.4 and mc=1.65 compared to the HERAPDF1.0 Adding in the charm data barely changes the PDFs.. m_c clearly affects only the fit to the charm data- and the predictions for F2c are almost the same whether the F2c data is included in the fit or not.

F2c predictions RTVFN scheme for mc=1.4 and 1.65 In the case that the F2c data is included in the fit Note for Mc=1.65 the gluon is larger at small x but the F2c prediction is suppressed

Has the F2charm data improved the uncertainties on the PDFs? Can only compare experimental red bands. I can see a marginal decrease in the size of the cbar uncertainty- encouraging HERAPDF1.0HERAPDF1.0 plus F2charm It is a separate question whether the input of this data restricts parametrisation dependence- much more work! However it is clear that the model uncertainty due to variation of charm mass can be reduced

FFN fits in QCDNUM17beta.06 CANNOT deal with CC data – no readily usable NLO coefficient functions for F2 or xF3, and this DOES matter because although the scale is high ZMVFN cannot be used sensibly- the problem is that the process W+c →s is missing and no coefficient function is making up for this! SO leave CC data out: 633 data points down to 565 Fit σ NC e+ (379), NCe-(145) and F2c (41) Hence FIX valence parameters. This leaves 6 parameters free from our standard 10 Parameter HERAPDF1.0 fit Make the heavy quark scale Q2+4mc2, best fits with mc=1.4 GeV χ2/ndp =567/565 for HERA_I combined 515/524 for F2c 52/41

The FFN/RTVFN change in the gluon is extreme.. I’ve seen it before but not so strikingly (Note the change in the Sea is just the lack of a charm PDF) Note: a Δχ2 improvement of 17 comes from adding an extra parameter to the gluon PDF→Dg ≠ 0, hence 7 free parameters χ2/ndp =550/565 for HERA_I combined 502/524 for F2c 48/41 No other extra Sea/glue parameters help

RTVFN mc=1.4FFN mc=1.4 Compare RTVFN and FFN with same mass mc=1.4 FFN is relatively suppressed (for Q2> 4 )and thus doesn’t NEED mc=1.65 to suppress it to get a good fit

RTVFN best fit mc=1.65 FFN best fit mc=1.4 scale Q2+4mc2, dg≠0 Compare RTVFN mc=1.65 and FFN mc=1.4 These are more similar, hence each give the best fit for their schemes

Can also see that contrary to RTVFN mc=1.4 gives a much better fit than mc=1.65: χ2/ndp =816/565 for HERA_I combined 546/524 for F2c 271/41 mc=1.4 χ2/ndp =550/565 for HERA_I combined 502/524 for F2c 48/41

Compare FFN predictions with different Mc FFN mc=1.4FFN mc=1.65 (this is now TOO suprresed)

Can do fits with heavy quark scale Q2 not Q2+4mc2- little difference to PDFs or to chisq χ2/ndp =552/565 for HERA_I combined 503/524 for F2c 49/41 The scales barely change the PDFs- or the F2c predictions

Compare F2c FFn with scales Q2+4mc2 and Q2 Slight difference at lower Q2 FFN scale Q2+4mc2FFN scale Q2

And we can cut out data Q2 > down to 502 data points- no significant difference to fit parameters χ2/ndp = 491/502 for HERA_I combined 434/461 for F2c 48/41 NOT illustrated Perhaps the extraction of F2c needs an FFN fit which does Not already have F2c data in it? This is illustrated here χ2/ndp = 500/524 Makes only a small difference to the PDFs whether F2c is fitted or NOT

So various FFN versions of the HERAPDF are available What do the Heavy flavour people want? Mc=1.4, scale=Q2+4mc2, with F2c NOT included in the fit is my guess And in what format do you need it? –Tables of PDF central values? (LHAPDF takes a bit of time, and I doubt this is an official release?)

Comparing HERAPDF1.0 to Tevatron data I am always predicting for the LHC, why not for Tevatron Run-II These are the predictions for W+/W- and Z, of course W+/W- are mirror images. It is compared to MSTW08. The Z0 data are actually closer to the HERAPDF - BUT NOTE although data errors are ~1.5% that’s not counting the normalisation of 6%, so both PDFs describe it fine.

Here are the W and the lepton asymmetries (ptlep> 20 GeV as for D0) YES we do differ from MSTW08, particularly for the lepton asymmetry However MSTW08 do NOT fit the D0 data so well either (nor high Pt CDFdata) Conclusion: W –asymmetry OK, not clear it would reduce our PDF errors much - lepton asymmetry problematic for everyone, but our result does look anomalous MSTW also has a fairly large error ~0.03 CTEQ66 similar D0 data illustrated CDF data illustrated

EXTRAS The point of this was just to check how we compare Rather than attempting to input TEV data to our fit – and it’s debatable if we would get much advantage- I want to see how we are doing after we get HERA-II data in the fit. Then I suspect there would be little extra to be gained from the TEV data

Further Z/W ratios for the Tevatron