Reconstructing low mass boosted A bb at LHCb

Transcription

1 Summer student report - CERN 2014 Summer student: Petar Bokan Supervisor: Victor Coco Reconstructing low mass boosted A bb at LHCb ABSTRACT: LHCb has the ability to trigger low mass objects with high efficiency. Some theoretical models predict low mass resonances decaying to. Mass Drop Tagger + Filtering procedure can be used to reconstruct these resonances in boosted regime. This report presents work on Mass Drop Tagger + Filtering parameters optimization in the LHCb kinematic regime. The parameters have been optimized by taking into account signal and background efficiencies and di jet mass resolution.

2 1 INTRODUCTION Quarks and gluons can t be detected as free objects. Almost immediately after being produced, a quark or gluon fragments and hadronises, leading to collimated spray of energetic hadrons a jet. Jets are obvious structures when one looks at an event display, and by measuring their energy and direction one can get close to the idea of the original parton. Jets are used for a few decades in particle physics already. There is no unique definition, but instead there is many jet algorithms. We would like for a jet to have the same properties as initial parton, starting with momentum and energy. This is much easier said than done. In real events, when quarks and gluons are produced, they can radiate additional gluons, there can be gluon splitting, or some other QCD processes. For this reason, jets can't be too small, because in that case jet will not represent original parton and this rises whole new problem. If two particles are too close in sence of their angular splitting, their products will be reconstructed as one jet. This would change signature of dominant process in event and it would lead to wrong conclusions. Solution lies in jet substructure. If those jets which were reconstructed as a jet coming from two primary event products can be identified, than the problem would be solved [1]. Jets coming from two primary event products are not the only one with substructure. As mentioned, gluon emission or gluon splitting can also give substructure to jets. Best case for analyzing substructure tools is hadronic decay of some heavy particle, because than we will be able to compare substructure results with our physical expectations. Substructure procedures are also used do discriminate underlying events, pileup, uncorrelated showering and hadronization products [2]. Jets wits substructure are often called fat jets. 1.1 Substructure on LHCb Previous work on LHCb, especially search for exotic massive long lived particles [3] gave us a reason to consider jet substructure. Search for exotic massive long lived particles in low mass region showed low efficiency in reconstructing resolved jets. Comparison of efficiencies for, where is some scalar particle, being reconstructed as resolved case (two jets) is given in table (Fig ). 35 GeV 25 GeV 15 GeV efficiency 60% 45% 7% Fig : Efficiency of reconstructing resolved jets for respect to different values of. in Angular separation (distance in space - ) between two resolved jets for di-jet event becomes smaller for lower masses, so only one jet can be

3 reconstructed (Fig 1.1.2). To increase efficiencies in Fig 1.1.1, one of the ways is to consider jet substructure, to identify those di-jet events that were reconstructed as one jet. Fig : between two resolved jets for di-jet event for different values. 2 DEFINITIONS 2.1 Mass Drop Tagger There is many tools which were developed to evaluate substructure. One of these tools is Mass Drop Tagger (MDT) which was specially designed for this particular case where only one jet is reconstructed for di-jet event. MDT is designed to be used with jets found by the Cambridge/Aachen algorithm (C/A) [3]. If we have jet labeled j and we wont to perform Mass Drop Tagger we have to do next [2]: 1. Brake jet j into two subjets by undoing its last step of C/A clustering. Label them j 1 and j 2 so that. 2. If and if splitting is not to asymmetric ( ) than j is tagged jet. 3. Otherwise redefine j to be equal to j 1 and go back to step 1 (unless consists of just a single particle, in which case the original jet is deemed untagged.

4 Parameter shouldn t influence results if it s not too small, and should be the most important parameter for Mass Drop Tagger. Standard choices are and (range). 2.2 Filtering Filtering is another procedure used to evaluate background. It is often used in combination with Mass Drop Tagger (Fig ). If is angular separation between the two prongs of the jet after tagging. The tagged jet was then reclustered with radius only its hardest prongs are kept. Alternatively we could keep only filtered jets that have more than some fraction of jet's transverse momenta (10%). This should also be investigated. and Fig : Mass Drop Tagger + Filtering 3 ANALYSIS 3.1 Process and boosted regime The goal of this analysis is to find range of parameters that can be used for Mass Drop Tagger + Filtering procedure that match conditions required by LHCb experiment. Signal MC samples that were used are different combinations of for process (Fig ). If, than will be boosted, and angular separation between two b quarks will be smaller than jet radius, so only one jet will be reconstructed for this process. Also, these samples are in accordance with MDT + Filtering procedure. Fig : Different combinations of To estimate boosted regime for individual ( how angular separation of b quarks depends on ) combination it s best to see distribution. To avoid dependence of this

5 analysis on initial distribution different bins are defined and most of the results will be presented in these bins. Since this analysis is about MDT + Filtering procedure we chose this way, but in real physical analysis only cut on minimum value would be applied. Bins are defined as: ( ) ( ) ( ) ( ) ( ) ( ). (3.1.1) The dependence of angular separation of b quarks on distribution (Fig ) is shown for three ( ) combinations, and bins are illustrated in (a) together with illustrative two fat jet radius choices, (red) and (black). (a) (b) Fig : Correlation between angular separation of b quarks and ( ) combinations. (c) distribution for three

6 These lines (red and black) do not represent fat jet radius literally, but they can give us some hints about optimal fat jet radius. In standard analysis usually there would be some cut which would define boosted regime ( ) and that value would determine optimal jet radius. The optimal value is different for different ( ) combinations, as showed in Fig Only data in LHCb acceptance are considered, so background this condition is applied for fat jets. cut is applied. For 3.2 Mass Drop Tagger parameter combinations In the table (Fig ) chosen combinations of parameters for Mass Drop Tagger procedure are shown. In the future it would be advised to check results for parameters in interval. 3.3 Filtering jet radius Fig : Mass Drop Tagger parameters Two values of filtering jet radius are used: and. These values perform differently, so more detailed study would be desirable. The optimization of filtering procedure is also connected to choice of criteria for number of filtered jets that are kept after filtering process. Filtered jets are recombined in filtered fat jet after filtering procedure, and mass distribution of filtered fat jet is relevant distribution for evaluation of filtering procedure. Angular separation between two subjets (after applying Mass Drop Tagger) depends only on fat jet radius (Fig ).

7 Fig : Angular separation between two subjets after applying Mass Drop Tagger Different number of entries in two histograms (Fig ) is only because more fat jets are reconstructed (in the same event) for smaller value of fat jet radius. On the other hand, angular separation between two filtered jets depends on both, for subjets (after applying Mass Drop Tagger) and choice of (Fig ). of Fig : Angular separation between two filtered jets after applying Mass Drop Tagger and Filtering. We can see that angular separation between filtered jets is influenced by. Choice is conditioned by efficiency of (filtered fat jet) mass distribution in region of Mother s mass. This is shown for fixed Mass Drop Tagger parameters in defined bins for fat jet

8 radius (Fig ). First few bins are not shown because of their low stats. Similar behavior is seen for all MDT parameters. Fig : Filtered fat jet ( ) mass distribution in defined bins for fixed values of Mass Drop Tagger parameters and for two different values of parameter. Two peaks can be observed. These plots are for, so we can conclude that fat jets in that mass region are fully reconstructed fat jets coming from. goes to two quarks ( ), so it s expected that these jets are jets with substructure and that they contain two hadrons. Since they have well defined substructure, their splitting is preferably symmetric and they will pass Mass Drop Tagger procedure. Better resolution and better efficiency for this second peak is criterion for optimization. Conclusion is that has better performance, especially in lower bins, as expected. For higher bins jets and subjets are more collimated and there is no difference between two values (look at the last histogram in Fig ) Other fat jets (that make first peak) can also be jets with substructure. Some accompanying processes, such as gluon radiation can lead to substructure and symmetrical separation. That s the reason why many of low mass jets will pass Mass Drop Tagger procedure also.

9 3.4 tagging Process in ideal case is process with two fat jets, one for each, since is boosted. As explained, well reconstructed fat jet contains two hadrons. If we ask for angular separation between filtered jet and hadron to be less than we can say that those filtered jets, which satisfy this condition, come from quark. If two filtered jets from one fat jet satisfy this condition, than that fat jet comes from a decay of mother. Next figure (Fig ) shows effect of fake tagging on filtered fat jet mass distribution. We say fake because we didn t discuss issue of LHCb s liability to separate two hadrons. tagging discriminates jets which are coming from, for example. Fig : Filtered fat jet mass distribution before and after tagging. 4 RESULTS 4.1 What was calculated for signal? Optimization of parameters includes assessment of their performance in respect to good signal efficiency, signal quality and background rejection. For signal, properties that were used as relevant properties are: Most Probable Value (MPV) Center of highest bin in mass histogram, but we require for MPV to be higher than some value ( for ). Mass Window (MW) defined as smallest width mass window that contains of signal (entries). Important parameters of MW are Width, Top Edge and Low Edge. Mass Resolution (Relative Width) defined as half of width divided by MPV.

10 tagging allows us to have much better overview of MPV and MW results. These results are shown in bins which are defined by (3.1.1). Some of these (first few) bins (after applying tagging) have low stats, so MPV and MW properties are not well defined for them. We can see that Mass Resolution improves for high values, but the difference is not too significant, so it s not good to go to high for, because signal efficiency drops (shown in sec. 4.2). Fig MPV for listed values ( ) boosted regime.

11 Fig : Width of MW for listed values ( ) boosted regime. Fig : Relative Width of MW for listed values ( ) boosted regime.

12 tagged filtered fat jet mass distributions for one of the bins ( ) is shown in (Fig ) and it can illustrate decrease in efficiency for high values. Fig : Filtered fat jet mass distribution for listed values in one of the bins. 4.2 Signal and background efficiency First, we divide mass region into five mass windows: in. We define four different efficiencies in 8 ( bins (3.1.1)) diagrams: 1. Number of fat jets (in LHCb acceptance) in specified fat jet mass window (5 mass windows) and for specified (6 values), normalized to total number of fat jets in that fat jet mass window. Mass Drop Tagger does not influence this number (it s not yet applied). 2. Number of fat jets that pass Mass Drop Tagger procedure (in LHCb acceptance) in specified fat jet mass window (5 mass windows) and for specified (6 values), normalized to total number of fat jets in that fat jet mass window. 3. Number of filtered fat jets (in LHCb acceptance) in specified filtered fat jet mass window (5 mass windows) and for specified (6 values), normalized to total number of fat jets in the same fat jet mass window. 4. Number of tagged filtered fat jets (in LHCb acceptance) in specified filtered fat jet mass window (5 mass windows) and for specified (6 values), normalized to total number of fat jets in the same fat jet mass window.

13 QCD jets ( production) are used as background samples. Samples with different minimal hadron transverse momentum are used separately. In the next two figures four different efficiencies (listed before) are shown for bb20 (at least one hadron with in every event) and bb120 (at least one hadron with in every event). Filtering of background fat jets leads to migrations of jet s mass (filtered fat jet mass) to low values, so high values of number of filtered fat jets can be seen in first mass bin in Fig and Fig In Fig four efficiencies are shown for signal ( ). For this figure it s interesting to see what happens in mass region of (region indicated with arrows).

14 Fig : Four efficiencies for background (bb20) in defined bins.

15 Fig : Four efficiencies for background (bb120) in defined bins.

16 Fig : Four efficiencies for signal ( ) in defined bins.

17 Fig 4.2.4: Four efficiencies for signal ( ) in defined bins, but in 68% Mass Window, which is defined for filtered fat jet mass distribution after MDT + Filtering and after B-tagging.

18 Number of fat jets (Fig ) is not always the same in one bin because MW boundaries are not the same, not because of change in values. Also, number of fat jets is not the highest number in every bin, because of the migrations (to lower values) of fat jets mass after MDT + Filtering. Results shows, what we hoped for, significant background rejection, with satisfying signal quality. Optimization of values showed that and have the best performance, so it would be advised to check values in interval. Also, more detailed study of values is needed for specific kinematic regimes. References: [1] G.P. Salam, Towards Jetography, (2010) [arxiv: v2[hep.ph]]. [2] M. Dasgupta, A. Fregoso, S. Marzani and G.P. Salam, Towards an understanding of jet substructure, (2013) [arxiv: v2[hep-ph]]. [3] LHCb-ANA [4] J.M. Butterworth, A.R. Davison, M. Rubin and G.P. Salam, Phys. Rev. Lett. 100 (2008) [arxiv: [hep-ph]].