Cross sections for processes with mixed orderEW/orderQCD diagrams

[Frank Siegert]

When generating processes without fixing the Order_EW or Order_QCD of the diagrams, Sherpa will wrongly enhance the interference between diagrams of different Order_EW/Order_QCD. This is not a problem in any of our default setups, because there we restrict the Order_EW, but if one includes all (e.g. because of cuts which enhance the EW contributions such that they are needed), one will run into this bug.

Example: Sherpa's cross section for pp -> P b bar around the Z->b bbar peak doesn't reproduce the one in Madgraph if the following cuts are applied:

• pT(photon) > 200 GeV
• pT(b) > 50 GeV
• pT(bbar) > 50 GeV
• |eta(photon)| < 2.5
• |eta(b)| < 2.5
• |eta(bbar)| < 2.5
• dR(photon, b) > 0.6
• dR(photon, bbar) > 0.6
• 89<m_{b bbar}<93

Using cteq6l1 and a fixed fac/ren scale=mZ, Madgraph gives 0.007710 pb, while Sherpa gives 0.020742 pb. If I select the resonant diagrams only, i.e. pp -> P Z[b bbar] then Sherpa (0.005917 pb) and Madgraph (0.005869 pb) agree well.

[Frank Siegert]

Hmmm, I just notice that the Madgraph integration results seem to have huge errors, ~100%. So the numbers above can't be trusted. I'll try to find out how to improve that. Also I have to change the ren/fac to actually be fixed (.true. instead of F) and maybe just to be safe set the bwcutoff higher, like 150. Will post back with new results.

[Frank Siegert]

Even if I integrate to lower errors (and with actually fixed scales), I still see the same discrepancy. I've simplified it even further, by looking at only one subprocess, namely u ubar -> gamma b bbar (in Madgraph that's uu~>bb~a and u~u>bb~a). Furthermore I've tried to identify whether it's the cuts that may be wrong by removing them successively. But whatever I do, in all cases Madgraph's cross section is only ~40% of Sherpa's. Are there any other ideas where this difference could come from?

[Frank Siegert]

This is fixed on Sherpa SVN but will need to be thoroughly tested before it can be released.

[Frank Siegert]

Works fine in our benchmark, except for Amegic with MHV amplitudes (#88), disabled them for the time being, nobody is using them anyway.

[Frank Siegert]

Works fine with Amegic as well now (i.e. in release 1.2.2).

Note that the SCALES and COUPLINGS parameter syntax has been changed to accomodate the fixes described in this report. Check the manual for details.

[Stefan Hoeche]

Milestone old deleted