This first set of trigger studies is based on p11.08 WH MC and a recent run. I've been doing this work along with Jodi Wittlin @ BU.

Data

• Run 160694 (see run conditions database)
• Used the D0Analyze_x program to dump the online trigger decisions (p12.01)
• Trigger list 8.0 was used to take the data.
• SAM dataset name run_160694_raw (look under user gwatts)

Monte Carlo

• WH, W->en p11.08 file, no minbias. Higgs is 115 GeV/c2.
• Simulated with p12.01 version of TrigSim_x, and hand edited version the trigger list 8.0 (to make it work)
• SAM dataset name whenu_p1108.

Summary

    Slides 8/30/2002 (ppt, pdf)

All rejection studies were done on data hanging of 8.0's bit 9. At (10E30, 30E30) its rate is (84Hz, 240 Hz).

L1
L1 Triggers: CJT(2,5)CJT(3,3)CEM(1,10)
L1 Efficiency: 93%
L1 Rejection: 2.8 -- Rates: (30 Hz, 85 Hz)

L1 w/MET
L1 Triggers: CJT (2,5)CJT(3,3)CEM(1,10)MET(10)
L1 Efficiency: 90%
L1 Rejection: 5 -- Rates: (17 Hz, 48 Hz)

L2
L2 Triggers: l2jet(1,20)l2jet(2,10)
L2 Efficiency: 96%
L2 Rejection: 1.6 -- Combined Rates: (15 Hz, 43 Hz)

L1 rates of close to 30 Hz are probably ok. L2 output rates close to 20 Hz are probably ok. Also, overlaps (especially with triggers in other groups) are not taken into account. So we may be able to get away with a trigger list this at L1 even at luminosities as high as 30E30.

A scan of all the cuts for L1 and L2, graphically, is here:

The signal efficency is 90%, and the background rejection is 5.5. If you add the MET cut in you will see the signal efficency drop to 85% and the background rejection increase to 8.3

At L2, due to a number of problems, I was unable to attain electron rejection. Likely this is due to bug(s) in the trigger simulator or the way I'm running the trigger simulator.

What is next?

• Level 3 Study
• Trigger list 9.0 is coming in several weeks. There is (finally) a move afoot towards simplifying the L1 triggers. For example, there may be an unprescaled CEM(1,10) trigger. It would be our responsibility to make sure that L2 and L3 triggers are good enough to keep the rate down for this channel.

Signal And L1/L2/L3 Bits

The following shows the fraction of events that passed L1, L2, and L3 by L3 trigger bit number. Note that the bit numbers are 1 off from the table below. Here trigger bit 0 is actually the min_bias trigger bit. This data is written in the Trigger Eff column in the table below.

Level 1 only -- a very obvious pattern, which you can see reproduced in the L1 Only Eff column of the table below.

Raw L2 Rejection Numbers From Data

Below are the raw numbers. A description of the columns:

• L3 Bit Number -- The bit number in the trigger list. Note that the main trigger list page starts with zero, but that is a script runner too. Bit 1 is the first physics bit. I've done the numbering below to follow that.
• Trigger Name -- Name of the trigger in the trigger db. Click on it to get a DB dump of that trigger.
• L1 Only Eff -- The fraction of signal events passing the L1 requirement. I think the 0.9996 is actually a but -- it should be 1.0.
• Trig List Eff -- The fraction of signal events passing the L1, L2, and L3 requirment. This would be the expected efficiency of the signal in the 8.0 trigger list.
• L2 Jet Cut -- A single L2 Jet cut in GeV (single possible point). The jet cut was chosen to get as close as possible to a 95% eff on the signal.
• Eff relative Default -- The efficiency relative to the Trig List Eff for the signal (so you must multiply these two numbers to get the total signal efficiency).
• Rejection Factor -- How much data that was taken on that L3 bit would be rejected by this L2 cut. A rejection factor of 1 means no data rejected, 2 is half the data, etc.

If the Trigger List Efficiency is small, then that L1/L2/L3 combination has already cut out most of our signal, so that row can be basically ignored. A great row to look at (and perhaps the most meaningful one) is bit 9, which is a CEM(1,5)CJT(2,5) -- one EM tower above 5 GeV and 2 jet trigger towers above 5, and there is no l2 or l3 requirement.

Trigger Bit 9

L2 Jet Et distributions for signal:

And in Data:

And an acceptance and rejection plot as a function of a single jet et cut in L2.

Level 1

CEM and CJT are obvious cuts for this channel. Start with the CEM, as you'll only want to require one of those The data is is black, and the MC is red. The spike at 60 GeV is saturated L1 trigger towers:

A scan over various possible cut values shows a 10 GeV cut (of which a refset is available) is just fine:

So the CEM(1,10) is the obvious term to use here. Gives us a fairly large rejection.

Here is the jet tower distribution, data and MC. Again, the spike at 60 GeV is the saturated towers:

Here is a scan over the defined refsets, requiring three jet towers (yes three -- electrons are in the same set, and there is no emf cut). The 3 GeV refset looks good.

And in addition, looking for two towers:

So here, a but with the 5 GeV refset looks like the right thing to do. So the nex thing to do is combine the three cuts all in one:

The first is no cut, then CJT(2,5) then CJT(3,3), then CEM(1,10). Efficiency is about 93%, rejection is 2.8.

L1 Missing ET

The MET term is not yet available at L1, and probably won't be until fall of 2002. However, it does offer an impressive amount of rejection. Below is data and MC and the calculated MET at L1.

If we do the usual scan we find that with a 5% loss in efficiency, we have a 1.6 rejection factor:

Level 2

Below are the jet et distributions for all events in data and MC that pass CEM(1,10)CJT(2,5).

The next two plots show what happens to the MC and Data L2 Jet distributions as the L1 cuts are applied. The cuts are in the following order: No Cut, CJT(2,5), CJT(3,3), CEM(1,10)

MC:

Data:

However, if you look at the # jets in MC and data and the leading jet ET, you do see the following:

And a cut on the leading L2 Jet:

At 20 GeV we get rejection -- not quite a factor of 2. A bit more can be done by requiring 2 jets at 10 GeV as well. If we do that, our efficiency is 96%, and background rejection is about 1.6 in L2 alone.

I looked at L2 electrons, but at this time there are a number of problems.

Conclusion: L2 is not adding very much -- because L1 is doing it so well.