The STAR experiment

I'm concerned about an apparent deficit of g-g scattering events in our CDF Tune A Pythia samples.  I first noticed this deficit when I was looking at MC asymmetries for pi-minus production, where the difference between asymmetries from g-g and q-g events is extreme.  For reference, here is Vogelsang's prediction for the subprocess mixture for inclusive pi0 production:

Compare that to the following plot from Pythia for pi-plus production after combining partonic samples starting at 3_4:

If I add 2_3 GeV the gg contribution at low pT gets a boost, but I'm still seeing ~zero pions from g-g scattering at 8 GeV, where Vogelsang predicts 20% g-g. 

Renee suggested skipping the observables and just plotting the event partonic pT.  The turnover from gg to qg occurs at 3 GeV ... seems too low to me.  On the other hand, the mixture is ~20% g-g at 16 GeV, which seems about right.

One other tidbit for any Pythia tuners out there.  When I was running standalone Pythia to get the various partonic xsections for weighting purposes I saw a number of advisory warnings like

in this case, I was running the 5_7 sample.  The initial max value for the q-g xsec was 5.3392D-01, and the final measured xsec was 1.515D-01.  Not sure if it's significant.

Update 11-19-2007

I looked into the relationship between charged pion p_{T} and event partonic p_{T} in more detail.  Here are plot of charged pion multiplicity and mean pion p_{T} versus event p_{T}:

so g-g events actually produce more charged pions per event, but these pions are all at low p_{T}.  The second conclusion is reinforced by the PDF I've linked at the bottom of the page, which shows pion p_{T} spectra split by subprocess for a range of partonic p_{T} bins.  The g-g events have a much steeper slope.  Those plots are not normalized by the # of events per subprocess, so the conclusion about the number of pions per event is not immediately evident.

Update 11-27-2007

I tried manually rescaling the g-g scattering contribution to the PYTHIA asymmetries; it looks like increasing the g-g by a factor of 5 does a decent job of reproducing the NLO theoretical predictions:

no rescaling

multiply g-g by 3.0

multiply g-g by 5.0

References

Agreement Between "Polarized" Pythia and NLO pQCD
Updated Charged Pion A_{LL} predictions

We're using an afterburner framework that turns Pythia into something like a LO polarized event generator to study any biases introduced by our triggers on A_LL measurements.  This page compares the asymmetries generated by Pythia to theoretical predictions from GRSV.

Cuts

• MC and reco vz positions inside 60 cm
• |eta| of reco primary track < 1.0
• dca of associated global < 1.0
• fit points > 25
• select pions using geant ID 8 (pi-plus) or 9 (pi-minus)

No trigger requirements are imposed.

Weighting

I combined MC samples 3_4 through 55_65 using the following cross sections and event counts

Plots

First off, here are the asymmetries integrated over all subprocesses.  The left column is pi-plus, the right column is pi-minus.  The agreement for postive charges seems basically acceptable to me, but pi-minus is off by quite a lot:

If I restrict to g-g, q-g, and q-q subprocesses individually, the difference between the two is obviously in the q-g subprocess contribution (as expected):

It seems to me that the lack of agreement between our Monte Carlo and the theoretical predictions could be due in part to a lack of gg / over-abundance of qg in the subprocess mix.  Here are plots of the subprocess mix from Pythia:

compare that to the prediction from Vogelsang et al. for inclusive pi0 production, where gg is the primary contributor until ~3 GeV:

In particular, only the 3_4 GeV partonic sample has more g-g than q-g at any pion p_T.  I'm going to try including the 2_3 GeV sample in my studies and see if that bolsters the gg contribution.