Here I look at the L2JetHigh and JP1 cross sections after the fixes to the simulation to address the problems discussed here ...

For this study, I have also done a strict run-by-run data to simulation matching; only those runs which have data and simulation are used. The number of runs discarded when I impose the matching is small and I will add them back in when I am confident all other major discrepencies are taken care of.

The integrated luminosities are needed to normalize the corrected data yield to the calculated cross section. I find the integrated luminosity for the various samples by looking at the total number of BBCMB_Cat2 triggers which fired during the runs in the particular sample, multiplying by the prescale (100000) and then dividing by the BBC cross section (26.1 mb).  

As a consistency check, I have divided the cross sections and the components which go into the cross sections into RFF and FF samples.

Figure 1: This figure shows the correction factors, which are the ratios of truth level to detector level simulation, for the L2JetHigh triggers (Upper 4 pannels) and JP1 triggers (Lower 4 pannels). For the plots showing spectra, the red curves are the unbiased particle level simulation and the blue curves are the detector level simulation that went through BFC trigger filtering. For the ratio plot (lower right pannel in each group of 4) the black curve shows the correction factor for the entire sample, the blue curve shows the cf for the RFF part of the run and the red curve shows the cf for the FF part of the run. In each group of 4 pannels, the upper left shows the spectra for the full sample, the upper right shows the spectra for the RFF sample, and the lower left shows the spectra for the FF sample. 

Figure 2: This figure shows the response matrices for the L2JetHigh (Upper pannels) and JP1 (Lower Pannels) triggers. The response matrix shows the particle level dijet mass vs the detector level dijet mass for events which have matching particle level and detector level dijets. The upper left pannel shows the full sample, the upper right pannel shows the RFF sample, and the lower left pannel shows the FF sample.

Figure 3: This figure shows the raw Data (Upper Left), Full Reco Detector Level Simu (Upper Right), and Unbiased Particle Level Simu (Lower Left) yields for the Full, RFF, and FF parts of the run. The data curves are normalized by the integrated luminosity of the sample and the simu curves are normalized by the number of runs included. The lower left pannel shows the FF to RFF ratio with the black curve showing data, the red curve showing the unbiased simu sample, and the blue curve showing the full reco sample.  

Figure 4: This figure shows the cross section obtained using the correction factors in figure 1 for the L2JetHigh triggers compared with theory. The top pannel shows the spectra with the black curve being the data cross section for the full sample, the blue being the cross section for the RFF sample, the green being the cross section for the FF sample, and the red being the theory. The bottom pannel shows the data / theory ratio for the full sample (Black), the RFF sample (Blue), and the FF sample (Green). 

Figure 5: This figure is the same as figure 4 but now for the JP1 triggers.

The difference between the RFF and FF raw data seen in figure 3 as well as the difference in cross section between RFF and FF, especially at low mass, seen in figures 4 and 5 seems a bit concerning. I would expect the cross sections from RFF and FF to be statistically consistent, but I see clear seperation with the FF cross section smaller than the RFF cross section. To investigate this further, I have divided the full data sample into 7 different 10 day periods (actually 6 10 day periods and day 180) and plotted the raw data dijet spectrum for each period. The raw dijet yield for each period is normalized by the integrated luminosity for that period calculated as described at the top of the page.

Figure 6: This figure shows the raw data dijet mass spectra from 7 different periods during the run. Each spectrum has been normalized by the integrated luminosity for the runs which make up the given spectrum. The left pannel shows the spectra and the right pannels shows the ratio of the the largest value to the smallest value of the normalized dijet mass in each mass bin. 

I have also created a pdf file which compares the luminosity scaled data dijet mass spectrum for each 10 day period with the spectrum from the entire run period.

To investigate the apparent drop in the cross section as a function of time, I have plotted the ratios for several trigger quantities as a function of run. First I look at the ratio of the jet luminosity monitor (BBCMB-Cat2) to the L2JetHigh triggers. I then try to isolate the behaviour of the BBCMB-Cat2 and L2JetHigh triggers by comparing each to the ZDCMB and VPDMB trigger numbers. All numbers where obtained from the run log page.

Figure 7: This figure shows the BBCMB-Cat2 / L2JetHigh ratio for all data runs which have a matching simu run. The 7 ten day periods have been color coded. 

Figure 8: Same as figure 7 but now showing the BBCMB-Cat2 / ZDCMB ratio.

Figure 9: Same as figure 7 but now showing the BBCMB-Cat2 / VPDMB ratio.

Figure 10: Same as figure 7 but now showing the L2JetHigh / ZDCMB ratio.

Figure 11: Same as figure 7 but now showing the L2JetHigh / VPDMB ratio.

Table 1: This table shows the average value for each ratio in each 10 day period. The average is just the average position of the points, there is no event number weighting. NB, several of the day 120-129 plots have runs with a value of zero (these come from the production_Lo runs which are not included in my main analyses but were included in the run list used to grab the values used to populate the ratio plots) and these will artificially pull down the average a bit.  

Figure 7 shows some interesting features. The average ratio seems to rise over the course of the run, possibly in conjunction with a rise in instantaneous luminosity. However, the fill structure can clearly be seen (especially for earlier runs) and the behaviour within a fill is opposite to the overall trend, at the begining of the fill when luminosity is highest the ratio is low and then increases as the luminosity in the fill decreases. One hypothesis is that the clear fill structure could be a result of large backgrounds in the early part of the fill. To test this, I have made a plot of the BBCMB-Cat2 / Dijet ratio where the dijet numbers come from my analysis before many cuts are applied. The idea is that the two jet requirement will cut down on the amount of background events firing the L0 L2JetHigh trigger and the fill structure will be less evident.

Figure 12: This figure shows the BBCMB-Cat2 / Dijet ratio. Here, the number of dijets are obtained from my analysis with very minimal cuts. I require that two jets are present in the event, the event had a z vertex less than 90 cm, one of the jets points to a fired jet patch, and both jets fall within my fiducial volume (-0.8 < eta < 1.8).