Adding timing to the VELO

Transcription

1 Summer student project report: Adding timing to the VELO supervisor: Mark Williams Biljana Mitreska Cern Summer Student Internship from June 12 to August 4, 2017

2 Acknowledgements I would like to thank Mark Williams for the excellent supervision provided during this summer internship spent working with LHCb, on his detailed explanations, thoughts and comments on my work.

3 Contents 1 Introduction 3 2 Description of the VELO design Current VELO design VELO upgrade design-phase Super VELO Project details and goals Detector model Simulation procedure Event generation Reconstruction of the event Association of the PV to b Results and discussion 7 5 Conclusion 8 6 Future work 8 7 References 8 8 Appendix 9 Abstract The LHCb experiment is designed to perform high precision measurements of matterantimatter asymmetries and searches for rare and forbidden decays, with the aim of discovering new and unexpected particles and forces. In 2030 the LHC beam intensity will increase by a factor of 50 compared to current operations. This means increased samples of the particles we need to study, but it also presents experimental challenges. In particular, with current technology it becomes impossible to differentiate the many (>50) separate proton-proton collisions which occur for each bunch crossing. In this project a Monte Carlo simulation was developed to model the operation of a silicon pixel vertex detector surrounding the collision region at LHCb, under the conditions expected after 2030, after the second upgrade of the Vertex Locator(VELO).The main goal was studying the effect of adding 4D detectors which save high-precision timing information, in addition to the usual three spatial coordinates, as charged particles pass through them. With the additional information on the particle timing, it is possible to separately reconstruct the individual 50+ collisions, allowing the next generation of high-precision measurements to be made at the LHCb. 2

4 1 Introduction The LHCb experiment was designed to investigate the difference between matter and antimatter by studying decays of beauty and charm hadrons. It is part of the LHC(Large Hadron Collider) situated as part of CERN(European Center for Nuclear Research). Measuring properties of b and c decays and searching for rare decays can help understanding the Standard Model and as well looking beyond it. As the production of the b-mesons is concentrated in the forward direction the LHCb experiment was constructed as a single arm forward spectrometer which covers the pseudorapidity region of 2 < η < 5. The construction of the experiment begins with the Vertex Locator (VELO) which is situated in the region of the proton-proton interaction. The VELO is a silicon detector that provides reconstruction of particle trajectories passing through and at the same time distinguishing between primary and secondary vertices. The tracker system (which consists of three other detectors) reconstructs their trajectories and determines their momentum. The RICH (Ring Imaging Cherenkov) detectors provide particle identification using Cherenkov radiation. A large dipole magnet curves the paths of charged particles and helps identifying them. Their energy is measured by using an electromagnetic calorimeter. The last piece of the experiment is the muon system whose goal is to investigate muons in each event and measure their properties. As part of my summer student project the work concerns performance studies of the VELO in the upgrade of Run 5 when the HL(High Luminosity) LHC will be established so the VELO will be explained in more details. Figure 1: Schematic view of the LHCb experiment. 2 Description of the VELO design 2.1 Current VELO design The VELO plays an important role in the performance of the LHCb experiment mainly because of its role of locating the primary vertices along the beam line and the secondary vertices from the decays of long lived particles. The current design consists of 42 silicon modules arranged perpendicularly to the beam. The sensors have radial geometry and are divided in two halves that during beam time approach each other at on approximate distance of 8 mm from the beam. Charged particles are reconstructed as tracks by identifying the pattern of hits in the detector. Primary and secondary vertices, and intermediale states(e.g b hadrons) are then reconstructed using the track information from the event. The current VELO design will be able to operate up to 2019 when an upgrade-phase 1 is expected to happen which will take two years to install two years ( ). The upgrade is a result of increasing the beam luminosity for which detectors have to be constructed in order to survive high radiation damages and to improve their performance. 3

5 Figure 2: The current VELO design Figure 3: The upgraded design of a VELO, module layout and a front view. 2.2 VELO upgrade design-phase 1 Due to the increasing collision intensity, the VELO silicon strip detector will be replaced by a pixel detector(fig. 3). It will consist of two halves with 26 modules each placed at an exact z position chosen according to previous simulation studies so as the detector can achieve the highest possible tracking efficiency. Since the VELO is the primary vertex finding system of huge importance is the study of its performance. 2.3 Super VELO Since the start of working of the HL LHC(Run 5) the LHCb experiment will receive 50x increased beam intensity compared to Run 2 which will influence the VELO and its performance. The planned Super VELO is considered to retain the already constructed version from upgrade-phase 1 consisting of the same spatial layout. The research now is moved onto improving pixel resolution by adding timing resolution on each module at each inner and outer part of the detector. 3 Project details and goals This project is a study of the VELO performance by adding the timing information in addition to the current spatial coordinates. The so called 4D detector model is implemented in the silicon pixel model of the VELO. While particles are passing through modules a precise measurement is made of its spatial coordinates and the time resolution. The VELO is determining the primary and secondary vertex position, especially important for further analysis is the secondary vertex association with a primary vertex. With the current detector model one vertex is present, as the luminosity is higher during the first and finally the second upgrade the number of primary vertices per event is larger. At Run 5 around 50 primary vertices per event are expected which again is increasing the probability of mis-association of the secondary vertex(sv - further denoted as SV) with the true primary vertex(further denoted as PV). The percent of assigning the wrong PV is strongly related of whether we use only spatial information or timing + spatial. Motivated by this reasons a Monte Carlo simulation was developed in order to investigate the effect of timing on the Super VELO. 3.1 Detector model The simulation is using a silicon pixel detector model for the VELO. Each module has a dimension of 35 mm. The simulated modules comprise a 3x3 grid of 15x15 mm 2 sensors with the central region empty(see Fig. 4).The inner and outer rings are allowed to have different time resolutions Each module has a specified z position which corresponds to positions in the real detector(taken from upgrade-phase 1)(Fig. 4). 4

6 Figure 4: Sketch of the Super VELO design. 3.2 Simulation procedure Event generation The process starts with generating a number of primary vertices along the beam line(fig. 5). Each primary vertex is generated at a certain point of time and space(fig. 6,7). While choosing the right spatial and time components for the PV a random number out of a distribution was chosen for which the plots (Fig 5,6,7) has been made. Figure 5: Distribution of a number of PVs generated. Figure 6: Z position of PVs along beam line. Figure 7: Time position of PVs along the beam line. After all the primary vertices have been produced next is to generate tracks from each of them which correspond to charged particles passing through the detector. Each track has a certain geometry which means it is produced by randomly assigning values for the pseudorapidity (η) and the azimuthal angle (φ) according to the known physics-level distributions. The number of tracks per PV is shown on Fig.8. Figure 8: Number of tracks that come out of the PV. Figure 9: Lifetime of b meson for a number of events 5

7 The track generation is followed by a b hadron vertex generation. Since the b hadron is a decay product there is a need of a PV to be chosen as a parent. So the simulation choses one PV for which the b is generated. The simulation takes into account some properties of the SV as the lifetime(fig.9) of the particle, its decay length and x,y,z components of the momentum. The b is considered as the secondary vertex and from each b (one b is generated per event) 2 tracks are produced Reconstruction of the event Since we develop a detector model, according to its geometry the reconstruction of particles and their tracks is made. When tracks(from PV and SV) pass through the detector they hit the modules and the number of hits are registered by the detector. Not all tracks are reconstructed, the simulation takes into account all tracks that have more than 3 hits (hit more than 3 detector modules). As a result one module has been chosen to perform the hit map for all particles that pass through it (Fig. 10). From Fig.10 and Fig.11 it can bee seen that the inner part of the detector experiences larger amount of hits. Moving further from the beam the module has less hits. Figure 10: Hit map of a module of the Super VELO (x and y position of hits). Figure 11: Hit map with z and y position of hits shown. Each hit has a certain time resolution that is saved after a check is made of whether the hit is situated in the inner or outer part of the detector. This helps to calculate the time resolution of a PV and SV. The vertex time is calculated by a weighted average of track times for all tracks in PV. t vertex = ttrack 1 1 σ 2 track σ 2 track To get track times all hits from each track are included and a weighted average is performed since each hit has a different time resolution. 1 thit σhit t track = 2 1 (2) σhit 2 The track time uncertainty is calculated by using the hit uncertainties which are only the time resolutions of the inner or outer part of the detector for the reconstructed tracks. 1 σ track = ( 1 (3) ) σhit 2 By calculating the track uncertainty the vertex uncertainty is calculated. 1 σ vertex = ( 1 ) σtrack 2 These formulas are used for the time and time resolution of the PV and SV. Later on these values are used to assign a b meson to a single PV. By only using the detector time resolution as shown from the formulas the PV time uncertainty is calculated. (1) (4) 6

8 σ hit σ track σ pv (5) The smeared position of each PV and the b and the smeared momentum of the b is found. The smeared position values are used to calculate the impact parameter(ip) for all PVs in the simulation. Figure 12: IP(Impact parameter) for true PV. Figure 13: PV time resolution Association of the PV to b While a larger number of PVs are present it is hard differentiating the true b parent. For associating the SV to a PV two different methods are used. 1. The first method is using only the spatial information after reconstruction of particles which is the IP. The PV with minimum IP is chosen as the correct one. 2. The second method is using the timing information which consists of already calculated values for the PV and SV time and PV and SV resolutions. All that is implemented in the formula for the 2D distance approach when assigning a PV. The PV with minimum value of: Here δt = t SV t P V and σ t = (IP ) 2 σ 2 IP σ 2 pv + σ 2 sv. + (δt)2 σt 2. (6) t (ps) δ IP(mm) IP(mm) Figure 14: IP distribution for PVs in an event. Figure 15: δt as a function of IP for PVs in an event. 4 Results and discussion By using the two methods for assigning a PV was concluded that both methods give different results. The first method chooses the vertex with minimum IP. In this case a mismatch fraction of 15% is predicted. A reason why the second method is introduced is shown in Fig.(14 and 15). On the Fig.(14) the min IP is the peak situated near 0. On the other hand in the Fig.(15) two PVs(with the IP near 0) have the same minimum IP but according to timing information there 7

9 is one which is closer to the b in time. This is a simple illustration when the timing resolves the situation and enough to get a more efficient method of associating a PV. Since 15% is a concern of choosing the right PV improvement was gained by using the 2D approach as explained above. The results show that choosing the wrong SV parent is less than 5% depending on the time resolution in the outer part of the detector(fig.(16)). From Fig.16 almost the same performance of the detector is expected if different inner time resolution of the detector is used (on the plot comparison is made between 200 ps inner time resolution and no timing in the inner part of the module) Correct PV Incorrect PV b lifetime residual (ps) Figure 16: PV mismatch fraction as a function of the time resolution in the outer part of the detector. Figure 17: b lifetime residual for the correct and incorrect PV. Another part of the discussion would be comparison of the b lifetime when we assign the correct PV and an incorrect(fig. 17). It has been made by calculating the lifetime of the b meson residual (blifetime measured blifetime reconstructed ). As it was expected the distribution is broader for the incorrect PV. 5 Conclusion In this study a Monte Carlo simulation was conducted in order to study the performance of the Super VELO. The detector model was developed by using the previous design of the VELO (upgrade -phase 1) and allowing the detector to deliver time information for the location of primary and secondary vertex. By studying the association of the SV to a PV parent it was concluded that timing improves the detector performance by reducing the fraction of mismatch from 15% to < 5%. 6 Future work A possibility of future work in this area is accounting for detector with different pixel sizes as this work considers 55 µm. This would allow reducing the 5% mismatch. Improving the selection algorithm for a PV is a potential future work by using machine learning techniques for the process of selecting the true parent of the b hadron during the event as an alternative to the 2D approach used here. 7 References [1] M.Williams, "Upgrade of the LHCb VELO detector",(2016), 14th Topical Seminar on Innovative Particle and Radiation Detectors, JINST 12 (2017) no.01, C01020, doi: / /12/01/c01020 [2] LHCb Collaboration, LHCb VELO Upgrade Technical Design Report (2013),CERN/LHCb- TDR-13. 8

10 8 Appendix For producing the final results shown on Fig.16 Table 1. and Table 2. has been used. Out time resolution % mismatch σ(p) Table 1: PV mismatch fraction calculated by 2D approach for different time resolution in the outer detector part with uncertainties obtained. The inner part of detector has 200 ps time resolution. Out time resolution % mismatch σ(p) Table 2: PV mismatch fraction calculated by 2D approach for different time resolution in the outer detector part with uncertainties obtained. The inner part of detector has no time resolution. 9