|Search for Boosted Top Quarks in High Transverse Momentum Jets with 5.95 fb -1 of CDF Run II Data.|
Raz Alon, Ehud Duchovni and Gilad Perez (Weizmann Institute of Science)
Pekka Sinervo (University of Toronto), [Contact]
We present the preliminary results of a search for boosted top quarks in a sample of high transverse momentum jets observed in the CDF detector in a sample of 5.95 fb-1. We observe 103 candidate events in a sample where we require either two massive jets or one massive jet and high missing transverse energy, with an estimated background of 76±10(stat)+26-20(syst) events. We use these data to set an the upper limit on the production cross section for Standard Model ttbar events with at least one top quark produced with pT > 400 GeV/c of 54 fb at 95% confidence level (C.L.)
This note follows on CDF Note 10199  ) that documents a study of the highly boosted jets using a 5.95 fb-1 sample of jet data. In this note, we extend the study to look more closely at the potential signals for top quark production in this sample, focusing on top quarks produced with pT > 400 GeV/c.
Our data sample consists of events collected with the jet100 trigger and selected to have at least one jet with pT > 300 GeV/c and |&eta| < 0.7. We considered jets reconstructed with the Midpoint algorithm with cone sizes of 0.4, 0.7 and 1.0.
The high pT jet sample is dominated by the QCD production of light quarks and gluons at all pT scales. The relative rate of top quark pair production rises as one increases the minimum pT requirement. Without any suppression of other sources, ttbar production contributes approximately ~ 1.5% at the pT > 400 GeV/c. With light quark, bottom quark and gluon contributions suppressed by a factor of 250, ttbar production represents over 50% of the signal for objects with pT > 400 GeV/c.
The most recent NNLO calculation of the ttbar differential cross section  has been updated with the MSTW 2008 parton distribution functions and a top quark mass of mtop = 173 GeV/c2. The calculation itself includes next-to-leading-order (NLO) corrections to the leading-order diagrams along with next-to-next-to-leading-order (NNLO) soft-gluon corrections. No rapidity cut was placed on this cross section though the author believes this would have a neglible effect on the overall rate. The scale used is μ2 = pT2 +Mtop2.
This calculation for the pT distribution yields a total cross section of 8.15 pb and a cross section for pT > 400 GeV/c of 4.55+0.50-0.41 fb. Said another way, the fraction of top quarks produced with pT > 400 GeV/c is 5.58x10-4.
In our calculations of expected ttbar contributions, we will employ the Kidonakis and Vogt cross section of 4.55+0.50-0.41 fb for top quarks produced with pT > 400 GeV/c. With this cross section, the PYTHIA MC sample for this pT range has a sensitivity of 888 fb-1.
We used a data sample collected with an inclusive jet trigger with a nominal transverse energy threshold of 100 GeV. The data sample corresponds to 5.95 fb-1 of Run II data. The entire inclusive jet sample consisted of 75,764,270 events for 5.95 fb-1. This corresponds to an effective triggered cross section of 12.7 nb.
More details of the event selection are provided in CDF Note 10199 .
With the selection described above, we believe that the event sample is dominated by jets produced by QCD scattering. The requirements that each event have a high quality primary vertex and that the calorimeter energy deposition associated with the leading jet be confirmed with charged tracks or the presence of both EM and HAD energy essentially eliminates all potential physics backgrounds and instrumental effects. We place a maximim value on SMET > 10 to reject backgrounds from cosmic rays.
The only other significant source of events to this sample is top quark pair production. Although the rate of top quarks is also expected to be of order 5 fb for pT > 400 GeV/c, these events will be unusual in that they will produce a small number of events with two massive objects. We therefore have considered these events as a potentially significant contribution to the event sample at high jet mass where the QCD rates are expected to be significantly reduced.
Top quark production is dominantly a pair-production process (ttbar) with the transverse momentum of the top quark being approximately half the mass of the quark, but with a long tail to higher transverse momentum. It is this tail that in principle contributes to any analysis looking at very boosted objects.
In order to understand the nature of this process and its characteristics when we require a central, high pT jet in the event, we used a standard top quark MC sample with 4.75 million events. As noted earlier, in order to normalize this sample, we employ the top quark cross section prediction by Kidonakis and Vogt prediction for top quarks with pT > 400 GeV/c. There are 4041 ttbar events with at least one top quark with pT > 400 GeV/c.
Given our event selection starts with a high pT jet in the central region, we make the same requirements on the top quark MC sample. We observe 1608 jets in this MC sample, which corresponds to an observed cross section for jets meeting these requirements of 1.81±0.20 fb, where the uncertainty includes the statistical uncertainty of the MC sample (almost neglible) and the uncertainty on the top quark cross section (which dominates). We note that this is approximately 250 times smaller than the observed rate of such jets in the data. As there are 1390 events that are responsible for these jets, there are 218 events in this sample with two jets with pT > 400 GeV/c. This corresponds to an expected ttbar event rate in this sample of 1.56 fb.
We start with 5233 events with a high pT leading jet requirement of jet pT > 400 GeV/c and |&eta| < 0.7.
A simple strategy to detect the presence of ttbar production when one is searching for fully-hadronic ttbar decays is to use both candidate jets in an equivalent manner. We start from the observation, illustrated in the figure below, that for QCD dijet events the masses of the two leading jets are uncorrelated. We can therefore use the observed distribution in either mjet1 or mjet2 of events in the low jet mass peak (defined here to be 30 to 50 GeV/c2) relative to events in the top mass window of 140 to 210 GeV/c2 to estimate the QCD background in the signal region where both jet masses are between 140 and 210 GeV/c2.
We find that there are 320 events with both jets in the mass region 30 to 50 GeV/c2 (region A). We also find 89 events with mjet1 ∈ (140, 210) and mjet2 ∈ (30, 50) (region B). We find an equivalent number of events - 113 - in the region mjet2 ∈ (140, 210) and mjet1 ∈ (30, 50) (region C). With these data, we estimate the number of QCD background events in the signal region (region D) to be 31.4±4.8 (stat). We observe 61 events in the signal region. This calculation is summarized in the table below.
Applying the same selection to our ttbar MC sample, we find 515 events in the signal region out of the 1390 ttbar MC events that have a leading jet with pT > 400 GeV/c. If we use the sensitivity of the MC sample of 888 fb-1, we would expect to see
in the signal region.
In order to observe ttbar events where one top quark has decayed semileptonically, we turn to the sample of high pT jet events where a recoil jet has not been identified as a potential top quark candidate through its mass. In this case, there is a correlation in the signal events with high mjet1 and SMET , forming a signal region defined by mjet1 ∈ (140, 210) GeV/c2 and SMET ∈ (4, 10).
By assuming that mjet1 is independent of SMET , we can perform a calculation similar to that used in the two high mass jet case. We define region A to be the one with mjet1 ∈ (30, 50) and SMET ∈ (2, 3), region B as mjet1 ∈ (140, 210) and SMET ∈ (2, 3), region C to be mjet1 ∈ (30, 50) and SMET ∈ (4, 10) and region D to be the signal region. We find that there are 269 events in region A, 65 events in region B and 186 events in region C. With these event counts, we predict 44.9±9 (stat) events in region D (the signal region).
Applying the same selection to our ttbar MC sample, we find 343 events in the signal region out of the 1390 ttbar MC events that have a leading jet with pT > 400 GeV/c. If we use the the sensitivity of the MC sample of 631 fb-1, we would expect to see
from ttbar production in the signal region.
We observe 42 events in this signal region, consist with the background estimate and also consistent with the number of expected background and signal events. This calculation is summarized in the table below.
Combining the results of the two channels, we find 103 candidate events with an expected background from QCD jets of 76.4±10.5 events (the uncertainty is only statistical). The systematic uncertainty on the background rate is dominated by the uncertainty on the jet mass scale (see the next subsection), and results in a background estimate of 76±10 (stat)+26-20 (syst) events. The statistical significance of this result is modest, given the lack of any excess in the lepton+jets channel, and is even less if systematic uncertainties are taken into account. Given this relatively modest significance, we cannot claim observation of high pT top quark production.
However, we do find that we can still set interesting limits on top quark production using these two channels. First, we estimate the systematic uncertainties associated with this measurement.
The largest systematic uncertainties affecting this analysis arise from uncertainty on the jet mass scale. The other sources of uncertainty we have considered are the integrated luminosity in the sample and the uncertainty in the top quark acceptance due to the uncertainty in the jet energy scale and the top mass used to create the MC samples.
We estimate the effect of the uncertainty on the jet mass scale by shifting the upper mass window by ±11 GeV/c2 and observing how the QCD background estimate changes. This results in a systematic uncertainty of -24% to +35% on the combined background rate of 76 events.
The jet energy scale uncertainty results in a systematic uncertainty on the top quark acceptance, which we determine by shifting the jet pT scale by 3% (the efficiency is sensitive to the jet energy scale simply because, for example, an underestimate in the jet energy scale would drop the observed rate of events and vice-versa). The resulting change in the top quark acceptance is 24.5%, using the pT distribution from the Kidonakis and Vogt calculation.
We find that the acceptance has an uncertainty of only 0.3% arising from the uncertainty on the top quark mass used to create the MC samples.
Finally, we incorporate a systematic uncertainty on the integrated luminosity of ±6%.
Together, these result in overall systematic uncertainties on the total cross section limit of -37% and +45%.
In principle, we can use the event rates observed in each channel as independent observations and combine them using a maximum likelihood technique or other statistical procedure to estimate the signal rate. However, given that we expect comparable signal-to-noise and acceptance in each channel, we combine the total number of candidate events and total background rate and use these to set an upper limit on ttbar production for top quarks with pT > 400 GeV/c. We calculate the 95% C.L. limit, folding in the systematic uncertainties, using the pseudo-experiment calculation developed by T. Junk and implemented in mclimit.C .
The resulting upper limit, taking into account the efficiency of 0.212 and the integrated luminosity of 5.95 fb-1, is 54 fb at 95% C.L. on standard model ttbar production for top quark pT > 400 GeV/c. This is approximately an order of magnitude higher than the estimated Standard Model rate, and is limited by the QCD background rates.
We note that the similar calculation using either of the two channels alone set upper limits that are within 50% of this limit.
Finallty, we calculate the "expected limit" by using the background estimated from the data-driven technique and assuming an observation of ttbar events at the expected level of 5.75 events. The mclimit calculation yields an upper limit of 39 fb at 95% C.L., which is lower than the observed limit since we see an excess of events above the expected signal plus background in the data.
We present the first search for very high pT top quark production using data gathered with an inclusive jet trigger. We find a modest excess of events - 103 candidate events with an estimated background of 76+29-22 events - either in a configuration with two high pT jets each with masses between 140 and 210 GeV/c2 or where we observe one massive jet recoiling against a second jet with significant missing transverse energy.
We expect approximately 6 signal events from Standard Model top quark production. The data are not sufficiently significant to support a claim for observation of top quark production. However, we do set a 95% C.L. upper limit on the rate of top quark production for top quarks with pT > 400 GeV/c of 55 fb at 95% C.L.
|Total Number of Observed Events in Signal Region||103|
|Predicted Background from QCD Jets in Signal Region||76±10(stat)+26-20(syst)|
|Expected Number of ttbar Events in Signal Region||5.75±0.72|
|95% C.L. Upper Limit on SM Top Quark Production with pT > 400 GeV/c||54 fb|
|95% C.L. Expected Upper Limit||39 fb|
|Table of Uncertainties|
|Table of Event Count - All Hadronic|
|Table of Event Count - Semileptonic|
|Distribution of Leading Jet Mass, SMET < 4, cone R=1.0, QCD and ttbar MC|
|Distribution of Sub-Leading Jet Mass, SMET < 4, cone R=1.0, QCD and ttbar MC|
|Distribution of Leading Jet Mass, 4 ≤ SMET < 10, cone R=1.0, QCD and ttbar MC|
|Distribution of Sub-Leading Jet Mass, 4 ≤ SMET < 10, cone R=1.0, QCD and ttbar MC|
|Distribution of mjet2 vs mjet1, SMET < 4, cone R=1.0, Data|
|Distribution of mjet2 vs mjet1, SMET < 4, cone R=1.0, QCD MC|
|Distribution of mjet2 vs mjet1, SMET < 4, cone R=1.0, ttbar MC|
|Distribution of SMET vs mjet1, cone R=1.0, QCD MC (see here for similar blessed plot for data)|
|Distribution of SMET vs mjet1, cone R=1.0, ttbar MC (see here for similar blessed plot for data)|
|Distribution of SMET vs mjet1, 4 ≤ SMET < 10, cone R=1.0, Data|
|Distribution of SMET vs mjet1, 4 ≤ SMET < 10, cone R=1.0, QCD MC|
|Distribution of SMET vs mjet1, 4 ≤ SMET < 10, cone R=1.0, ttbar MC|