I already checked the tridents, and they don't have angular cuts at the egs5 level.
Any issue would be within the MadGraph cuts, but I can explore these.
On May 13, 2016, at 3:25 AM, Bradley T Yale <[log in to unmask]> wrote:
[log in to unmask]"><BadMollers.png>[log in to unmask]"><2pt3_mol_v3_moller_E.png>_______________________________________________I think I found the real problem with how the Moller generator was initially set up.
This one affects the generated distribution a LOT more than the RNG precision probably did, and explains the remaining strangeness in the generated distribution.
Looking at the egs5 Moller procedure, the angular cut was defined as:
abs(v(np)) > 0.010 radians
where v(np) is supposed to be theta.However, the variables u, v, and w in egs are actually directional cosines, p_x = p*u, p_y = p*v, and p_z = p*w.
So this means that in reality, the generator was saving Moller events such thatabs [ sin(theta)*sin(phi) ] > 0.010
which has a periodic nature to it. Plot this equation for some value of phi (or just think about it) and you'll see what was likely making these strange hills and gaps in the energy distribution - full-wave rectified Mollers!
The scattered beam simulation does correctly define theta though:
sqrt[ u^2 + v^2 ]
I made a moller_v3 procedure with this correction (still with a >10 mrad cut), and the comparison between before and after is shown.
The generated events now agree with the calculated cross section (XS curve is shown on the 'bad' plot), and no apparent missing events.
These will be run through recon to see if we can finally get good Moller agreement with data.
I'm also going to try changing the scattered beam energy cut shown in the software meeting to see if it fixes Tim's phi vs. energy discrepancy.
-Brad
Hps-analysis mailing list
[log in to unmask]
https://mailman.jlab.org/mailman/listinfo/hps-analysis