Home » Groups » STAR Protected

Run 9 200GeV Particle Vs Detector Jet Comparison

Updated on Wed, 2012-12-12 16:36. Originally created by pagebs on 2012-12-08 23:50.

A more complete particle vs detector level jet comparison ...

This page is meant to expand on studies started in this blog post. I have added several new plots which I hope will make the observed trends easier to interpret. I have also utilized a new simulation created by Pibero which places a particle level jet low pt threshold at 1.5 GeV instead of 3 GeV. There are some other changes in jet matching and neutral fraction cuts which are described below.

Note: Several of the plots I show refer to a particle level neutral fraction. This is an ill-defined quantity because at the particle level I do not know how much energy a particle has deposited in the calorimeter. The particle level neutral fraction I quote is obtained by taking the ratio of neutral particle pt for species I would expect to deposit most of their energy in the calorimeter over the total pt of the neutral species and charged species. The neutral particle I count are: pi0, eta, rho, K0, K0_S, K0_L, and gamma. The charged particles I count are: Pi+/-, Proton, K+/-, Sigma+/-, Delta+/-, Xi-, Omega-, and muon.

When matching particle level jets to detector level jets, the first step is to make sure the particle level and detector level vertex match. I currently require the z position of the particle and detector level vertices to be less than or equal to 2.0 cm appart to use the event.

Figure 1: The difference in z vertex between the detector and particle level events.

 
 
 
For each detector jet which passes the dijet conditions, I find the particle level jet which is closest in eta-phi space by finding the particle jet which minimizes delta_R = Sqrt[(etaDet-etaPar)^2 + (phiDet-phiPar)^2]. I say a particle level jet matches a detector level jet if delta_R < 0.5 .

Figure 2: This plot shows the geometric matching between the detector and particle level jets. The upper left pannel shows detector - particle level jet eta. The upper right pannel shows detector - particle level jet phi and the bottom pannel shows delta_R.

The detector level branch has a low jet pt cutoff of 5 GeV. The particle level branch originally had a low jet pt cutoff of 3 GeV but it was lowered to 1.5 GeV in the latest simulation iteration. I currently use the 1.5 GeV simulation because the particle level to detector level matching is better.

Figure 3: This figure shows the distributions of detector level jets which do not have a matching particle level jet. The top three pannels are for the simulation created with a 3 GeV particle jet pt cutoff and the bottom three pannels are for the simulation created with a 1.5 GeV particle jet pt cutoff.

I have made one additional change to the dijet selection process used in previous comparisons. I have removed the condition that barrel jets must have a neutral fraction less than 0.95 and have at least 0.5 GeV of track pt. Now jets in any part of the detector can have a neutral fraction of 1. I have added the condition that both jets in the dijet cannot have a neutral fraction of 1.

Figure 4: This figure shows the properties of dijet events in which both jets have a neutral fraction of 1.0. The top two pannels show the jet pts of the two jets. The bottom left pannel shows detector eta vs eta and the bottom right pannel shows physical eta vs eta. 

There are six different trigger categories a jet can be included in in my dijet analysis. The jets in each category can in theory have different behaviour as they move into the endcap, so it is important to look at the behaviour for each category separately.

JP2:

Figure 5: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the JP2 trigger category.

 
Description: The JP2 trigger category contains jets which fired the L2JetHigh trigger and satisfied the L2 monojet, dijet, or random condition and had a pt greater than or equal to 8.4 GeV. The L2JetHigh trigger required either a jet patch fired above JP-Th2 or two adjacent jet patches fired above JP-Th0.

Naive Expectation: There needs to be a large amount of neutral energy in the jet to pass the JP2 threshold. I would expect track pt to play a small role at low jet pt leading to a relatively flat detector vs particle pt ratio or difference across eta. In higher jet pt bins I would expect track pt to make up a larger component of the total jet pt so I would expect to see more deviation in the detector vs particle pt ratio and difference as jets move to higher eta and tracking is lost.


Observation: As seen in figure 5, the detector vs particle level jet pt ratio and difference are relatively flat across all values of jet detector eta. There is some downturn at high eta but it is slight. Page 7 of the following pdf shows the average detector and particle level neutral fraction as a function of jet detector eta. In the barrel region, the neutral fraction is large in the lowest pt bin and decreases in higher pt bins supporting the expectation that track pt would play a small role at low pt and a bigger role at high pt. This can also be seen on page 13 where I show the average track, barrel tower, and endcap tower pts in a jet. It is interesting to compare page 13 with page 17 showing the average number of tracks, barrel towers, and endcap towers in a jet. Even though the average track pt increases for higher jet pt bins, the average number of tracks remains almost constant.

The behaviour of the detector vs particle ratio and difference is roughly what I expected at low pt, a slight trend to more average particle than detector jet pt. At higher jet pt however, I don't see the strong deviations I expected. Page 13 of the pdf shows that for the 10-15 and 15-20 GeV jet pt bins, the average track pt falls in the endcap region as expected but is compensated by a steep rise in the average endcap tower pt. In the higher jet pt bins, the average track pt stays roughly constant over the jet eta range. This may be due to a "selection bias" which means that the jets we reconstruct in the endcap at high pt tend to be those for which we reconstruct almost the true pt as opposed to higher pt particle jets for which we reconstruct only some fraction of the pt. This hypothesis is supported by page 7 which shows the detector and particle level average neutral fractions moving closer to the same values in the endcap region for the higher pt bins. Support can also be found on page 9 where I plot the particle level jet pt as a function of eta in each detector level jet pt bin. In the higher pt bins, the distribution becomes narrower at high eta, indicating that more often, the particle level pt is closer to the detector level pt.

All JP2 trigger plots can be seen here.

JP1Lo:

Figure 6: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the JP1Lo trigger category.

 



Description: The JP1Lo trigger category contains jets which fired a jet patch above the JP1-Th1 threshold and had a pt less than 8.4 GeV. Note: since the jet pt is constrained to be less than 8.4 GeV, only one jet pt bin is shown.

Naive Expectation: The jet needs enough energy deposited in the towers to fire the JP1 trigger but the jet cannot have a pt of 8.4 GeV or above, so the track energy should be rather minimal. Since the track contribution should be small, I would expect the detector vs particle jet pt ratio and difference to be relatively flat as a function of jet eta.

Observation: The detector vs particle ratio and differences fall off somewhat as jets move into the endcap region. The drop off in the ratio seems to be more pronounced than in the JP2 case, but may be due to the smaller detector jet energy in the numerator. The magnitude of the detector-particle level jet pt difference in the endcap region is roughly what is seen in JP2, although the magnitude of the difference is greater in the barrel region in JP2 making the dropoff seem sharper in JP1Lo. Page 13 of the JP1Lo pdf also shows that the track and tower components contribute somewhat equally to the total pt of the jet in the barrel region and the track and barrel tower components drop in the endcap region while the endcap tower component increases. Comparing to page 13 of the JP2 pdf, we see that the average track pt is roughly the same between the two triggers whereas the tower components are larger in JP2 as expected.

All JP1Lo trigger plots can be seen here.

JP1Hi:

Figure 7: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the JP1Hi trigger category.

 
 
  
Description: The JP1Hi trigger category contains jets which fire a jet patch above the JP-Th1 threshold. In addition, no jet patch above the JP-Th2 threshold can have fired and the jet must have a pt greater than or equal to 8.4 GeV.

Naive Expectation: The JP1Hi trigger selects a bit of an odd sample of jets. The amount of energy deposited in the towers must be high enough to pass the JP-Th1 but at the same time no jet patch can be above the JP-TH2 threshold so the tower energy is pretty constrained for any total jet pt. This means that the track pt will contribute a significant amount to the total jet pt in all bins. Given the expected predominance of tracks, I would expect the detector vs particle ratio and differences to fall off significantly as jets move into the endcap.

Observation: The plots in figure 7 show the exact opposite of my naive expectation, there is very little or no drop off in the ratio or difference as jets move forward in eta. Looking at page 7 of the pdf, the rise in detector level neutral fraction in the endcap region is not nearly as large as is seen in either the JP2 or JP1Lo indicating that track pt contributes at roughly the same level in the endcap as it does in the barrel. This is confirmed by looking at page 13 which shows the average contributions from track and tower pt to total jet pt at detector level. The red squares which show the average track pt are basically flat as a function of eta. Note that on page 17, the average number of tracks in a jet does decrease slightly as the jet moves forward in eta.

It is my hypothesis that this behaviour is again due to a selection bias in that the jets we reconstruct in this trigger are predominantly those for which most of the track pt is reconstructed. If a large fraction of the track pt is not recovered, the jet does not show up in this trigger. For JP1Hi, it appears that this selection bias starts at the lowest pt bin, whereas for JP2 it doesn't set in until roughly the 20-25 detector jet pt bin where we can see the average track pt remaining pretty flat across eta. If this selection bias is behind the behaviour seen, then it becomes very important to understand reconstruction efficiency effects.

All JP1Hi trigger plots can be seen here.

UnJP2:

Figure 8: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the UnJP2 trigger category.

Description: The UnJP2 trigger category contains jets which were not included in either JP2, JP1Lo, or JP1Hi but are in a dijet where the other jet was included in JP2. 

Naive Expectation: For a jet not to pass any of the jet patch triggers, the amount of tower energy must be low so I expect track pt to be the dominant contribution to overall jet pt in all jet pt bins. Since the track component should be prominent, I would expect the detector vs particle ratio and difference to drop as jets move into the endcap.

Observation: Much like for the JP1Hi case, the actual results defy my naive expectations. There is some downturn of the ratio and difference in the 5-10 GeV jet pt bin in figure 8, but the ratio and difference stay flat for the other pt bins. Page 13 of the UnJP2 pdf shows the selection bias behaviour seen before: the average track pt remains flat as a function of eta for all but the lowest jet pt bin. We also see on page seven that the average detector and particle level neutral fractions agree quite well across eta for all but the lowest pt bin where the detector level neutral fraction increases steeply in the endcap region. We also see that the average neutral fraction is much lower for UnJP2 than any of the triggered categories as expected.

All UnJP2 trigger plots can be seen here.

UnJP1Lo:

Figure 9: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the UnJP1Lo trigger category.

Description: The UnJP1Lo trigger category contains jets which were not included in either JP2, JP1Lo, or JP1Hi but are in a dijet where the other jet was included in JP1Lo. Note: because JP1Lo jets must have a pt less than 8.4 Gev, any jets which appear in the high pt bins of the UnJP1Lo trigger category must have a large pt imbalance.

Naive Expectation: Because these jets are untriggered and the only difference between untriggered categories is what the other jet in the dijet was, I would expect the behaviour of the UnJP1Lo jets to be the same as the UnJP2 jets.

Observation: Comparing figures 8 and 9 we see that the detector vs particle ratio and difference both show a slight downturn as jets enter the endcap in the first pt bin for both UnJP2 and UnJP1Lo which disapears in highter pt bins. However, both the difference and ratio plots for the UnJP2 category show a constant negative offset in the first two pt bins which is not present in the UnJP1Lo plots. On page 7 of the pdf we see that the rise in average detector level neutral fraction as jets move into the endcap is present in the first two pt bins (statistics are poor in the remaining bins) as opposed to only the first pt bin in the UnJP2 plots. We also see that in the 10-15 pt bin the average particle level neutral fraction more closely tracks the detector level neutral fraction in the endcap region. Page 13 also shows that the downturn of average track pt persists to the 10-15 GeV pt bin.

All UnJP1Lo trigger plots can be seen here.

UnJP1Hi:

Figure 10: This figure shows the detector / particle level jet pt ratio for 6 different detector level jet pt bins (top six pannels) and the detector - particle level jet pt difference for the same 6 pt bins (bottom six pannels). This is for the UnJP1Hi trigger category.


Description: The UnJP1Hi trigger category contains jets which were not included in either JP2, JP1Lo, or JP1Hi but are in a dijet where the other jet was included in JP1Hi.

Naive Expectation: I expect the behaviour of the UnJP1Hi jets to be the same as the other untriggered categories.

Observation: Like the other two untriggered categories, the UnJP1Hi detector vs particle ratio and difference plots show slight downturns as jets enter the endcap region in the first pt bin. These features disapear in the subsequent pt bins. As with the UnJP2 case, there is a constant negative offset in the difference and ratio across eta but it appears smaller in the UnJP1Hi category and is largely gone by the second pt bin. Looking at page 7 of the pdf, we see that in the first pt bin the detector neutral fraction average rises in the endcap region while the average particle neutral fraction stays flat. By the second pt bin, both particle and detector neutral fraction stay flat and consistent across all eta. This is consistent to what was seen in the UnJP2 category although the actual averages shift slightly between the two categories. Page 13 shows a similar trend where we see the average track pt component dip in the first pt bin and then flatten for subsequent pt bins like it does for the UnJP2 case.

All UnJP1Hi trigger plots can be seen here.

In addition to what is presented above, I have created a pdf which shows all trigger categories combined. Caution should be used when interpreting these plots however because event weighting can make trends appear where they may not be real. For example, on page seven, we see the average particle level pt increase significantly as jets move into the endcap for the first three pt bins, but that trend is much softer for many or all of the individual trigger plots. The difference comes from the fact that some some of the triggers have high particle neutral fractions and are highly weighted.

General Discussion:

I have presented what is hopfully a comprehensive overview of the detector level vs particle level jet pt behaviour and offered some ideas as to what could be the causes behind some of the trends seen. I want to make a few closing points to reiterate these ideas and point out some things which could be important moving forward.

  • As I described in my naive expectation sections, I originally thought that the triggers and pt regions which should be dominated by track pt would show the most difference between detector and particle level pt in the endcap region. In many cases, the opposite turned out to be true. Its my hypothesis that we are seeing a 'selection bias' in these cases. For example, say we reconstruct a jet in the 15-20 GeV bin of the JP1Hi trigger, this jet should be dominated by charged energy. It is much more likely that for this jet the charged energy was reconstructed properly and the detector level pt closely matches the particle level pt rather than the alternative which would be the jet we see is actually a much higher pt jet where a large fraction of the track pt is lost. The jets for which a large fraction of the track pt is lost end up in a different pt bin, a different trigger category, or perhaps are not reconstructed at all. This seems to be supported by the plots on pages 13 of the pdfs where we see the average track energy remaining flat across the full eta range in many cases.


  • If the above is true, it will be important to understand the various efficiencies and the underlying particle jet behaviour.


  • The 7th page of the trigger specific pdfs show the average detector and particle level neutral fractions of the jets as a function of jet detector eta. We see large differences in average neutral fraction as a function of pt bin as well as trigger. These trends seem to largly agree with what one would expect. In some cases we also see a rise in average particle jet pt as jets move into the endcap. It is my understanding that neutral fraction biases like this have been folded into the trigger and reco bias in the past. Are these effects large enough that we have to treat them in some other way?

 

Groups: