Performance of the RICH Detectors of LHCb

Available online at www.sciencedirect.com

Physics Procedia 37 (2012) 613 – 620

TIPP 2011 - Technology and Instrumentation for Particle Physics 2011

Performance of the RICH detectors of LHCb A. Papanestis∗ STFC - RAL, Didcot, OX11 0QX, UK

Abstract Hadron identification, in particular the ability to distinguish charged kaons and pions, is crucial to many of LHCb core physics analyses. LHCb Ring Imaging CHerenkov (RICH) detector fulfils this role by providing charged particle identification in the momentum range between 1 and 100 GeV/c. The calibration and monitoring of the RICH detectors is achieved using samples of D∗ , KS0 , Λ and φ events, which are plentiful in the data and can be cleanly isolated through their decay kinematics. The particle identification performance of the LHCb RICH detectors, measured in data taken during the 2010 and 2011 LHC runs, will be presented along with the strategy for aligning and calibrating the detector. Finally this performance will be placed in context by highlighting the impact of the RICH performance on some of LHCb benchmark analyses. c 2012 2011Published CERN, for benefitB.V. of the LHCb Collaboration. Published Elsevier BV.ofSelection and/orcommittee peer-reviewfor © bythe Elsevier Selection and/or peer review under by responsibility the organizing under11. responsibility of the organizing committee for TIPP 2011. TIPP Keywords: LHCb experiment, RICH detectors

1. Introduction LHCb is an experiment at the Large Hadron Collider dedicated to the study of CP violation and rare decays of heavy flavours [1]. The physics goals of LHCb and the expectations of some key physics measurements are described in Ref. [2]. The LHCb RICH system has been designed to identify charged particles, especially hadrons, in the momentum range of 1–100 GeV/c. This is achieved with two RICH detectors and three radiators (solid silica aerogel and C4 F10 in RICH 1, CF4 in RICH 2). A detailed description of the LHCb RICH system can be found in Ref. [3]. Particle identification is crucial in a flavour physics experiment in a hadron collider for the reduction of the combinatorial background of multi-body final states. Correctly identifying the kaons, typically present in heavy flavour decays, helps in the selection of the charged tracks of interest and reduces the number of incorrect combinations. Particle identification also helps to differentiate otherwise identical final states. An example of this is for the two-body hadronic decays, B → h+ h− . In this particular case there are many contributions, as illustrated in Fig. 1, including B0 → π+ π− , B0s → K + K − , and other decay modes of the B0 , B0s and Λb , which overlap in the invariant-mass plot. These different decays will have different CP ∗ On

behalf of the LHCb RICH collaboration Email address: [email protected] (A. Papanestis )

1875-3892 © 2012 Published by Elsevier B.V. Selection and/or peer review under responsibility of the organizing committee for TIPP 11. doi:10.1016/j.phpro.2012.02.413

614

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

Fig. 1. Invariant mass distribution for B → h+ h− decays in the Monte Carlo simulation, where h indicates a charged

hadron, showing the contribution from different B hadronic decay modes.

asymmetries and for a detailed study of CP violation it is important to separate each final state. This is possible using information from the RICH detectors that identify pions and kaons with high accuracy. The two LHCb RICH detectors have very similar geometry. Two segmented and tilted spherical mirrors focus the Cherenkov cone into a ring. Two also segmented planar mirrors are used to reflect the photons towards the photon detectors resulting in a more compact detector. The photon detectors are the pixel Hybrid Photon Detectors (HPD) that combine vacuum tube technology with silicon pixel sensor readout [4], specially developed with industry for the LHCb RICH detectors. The complex geometry of the detectors means that alignment and calibration of the hardware and evaluation of the particle identification algorithms is very important in order to achieve the best possible results. 2. Calibration and alignment The photon detectors of the RICH system operate in the fringe field of the LHCb dipole magnet. The presence of the magnetic field affects the electrostatic focusing of the photon detectors and results in image distortions that must be corrected for. The procedure for the correction of the image distortions uses a test pattern that is projected on the photo-detector plane both with the magnetic field off and on. Dedicated algorithms are used to identify the light spots and extract the parameters for the image correction in the presence of magnetic field (see Fig. 2).

1000

μcore = 0.016± 0.006 mm

800

σcore = 0.498± 0.006 mm

After Calibration Before Calibration

600

fraccore = 87%

400 200 0

-2

0

2

Residual / mm

Fig. 2. The spacial residuals, demonstrating the resolution with which the Magnetic Distortion Mapper (MDM) light spots are identified. The plot shows the distance from the measured light-spot centre, projected on the photo-cathode plane, to the nearest MDM grid-point. Red circles are before and blue triangles after calibration.

615

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

In order to accurately reconstruct the Cherenkov angle of the detected photons, the various components of the RICH detectors must be aligned and the whole RICH system must be aligned with respect to the LHCb tracking system. The components that require alignment are the silicon sensors inside the photon detectors, the photon detectors themselves and the mirror system. There are 4 spherical and 16 planar mirrors in RICH 1 and 56 spherical and 40 planar mirrors in RICH 2. The silicon sensors inside the photon detectors can be aligned independently of the tracking. Once enough events are accumulated an image of the circular photo-cathode is formed on the silicon pixel detector and the centre of the circle is identified. It has been observed that the photo-cathode image on the silicon sensors is not stable in time and it can move with time-scales of a few hours. A procedure that uses a Sobel [5] filter to enhance the edges of the photo-cathode image has been developed and it is automatically applied at every run. The results are stored in the Conditions Database [6] so that the alignment of the silicon sensors is used in the subsequent reconstruction of the LHCb events. All the other components of the RICH detectors are aligned using information from the tracking. There are two angles that can be calculated for every photon-track combination. The angle θ is the angle away from the track that defines the Cherenkov cone and the azimuth angle φ is the angle around the track. By selecting only high momentum tracks with a known (saturated) Cherenkov angle it is possible to plot the reconstructed minus the expected Cherenkov angle (ΔθC = θrec − θexp ) with respect to the azimuth angle φ. In an aligned system (where the projection of the track through the imaging system is at the centre of the Cherenkov ring) there is no dependence of the Cherenkov angle θC on the angle φ, so the distribution is flat (Fig. 3 right). Any misalignment in the system can be seen as a dependence of the Cherenkov angle θC on the angle φ, in which case the plot shows a cosine type curve (Fig. 3 left). Fitting the cosine curve with a function of the form: ΔθC = Acos(φ) + Bsin(φ) can extract the misalignment parameters A and B which are proportional to mirror rotations around the y and x axis respectively. For the alignment of individual mirror segments special isolation criteria are applied on the photons to determine their origin and a separate alignment histogram is produced for every spherical-flat mirror combination. The orientation parameters of the individual mirror segments are extracted in a subsequent step using a minimization algorithm as described in [7] and [8]. The alignment is implemented in software. It is possible to re-align the mirrors if required, however, it is an operation that can only be performed during a long stop of the experiment. Δθ v φ A Side RICH2

Entries 1.161066e+07 χ2 / ndf 2018 / 17 SinCoeff -5.57e-05 ± 1.322eCosCoeff -0.0005054 ± 1.696e Shift 0.0003088 ± 1.056e-2000

Δθ v φ A Side RICH2

Entries 5.981273e+07 χ2 / ndf 231.9 / 17 SinCoeff -2.366e-05 ± 8.678e 9000 CosCoeff 5.991e-06 ± 1.085eShift -1.439e-05 ± 6.913e

0.005

Δθ [rad]

Δθ [rad]

0.005 0.004 0.003

0.004

8500

0.003 1800

0.002 0.001

1600

0 1400

8000 0.002 7500

0.001 0

7000

-0.001

6500

-0.002

1200 -0.002

6000

-0.003

-0.003

5500

-0.001

1000 -0.004 -0.005 0

-0.004 1

2

3

4

5

6 φ [rad]

-0.005 0

5000 1

2

3

4

5

6 φ [rad]

Fig. 3. Δθc plotted as a function of φ (defined in the text) and fitted with Acos(φ) + Bsin(φ), for one side of the RICH 2 detector. The left-hand plot is prior to alignment, and shows a dependency of the angle θc on the angle φ. The right-hand plot is after the alignment correction, and Δθc is uniform in φ.

616

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

3. Performance 3.1. Cherenkov angle resolution The resolution of the Cherenkov angle can be measured by selecting tracks above a certain momentum that produce photons at a known Cherenkov angle. A histogram showing the reconstructed minus the expected angle is shown in Fig. 4 for the gas radiators in RICH 1 and RICH 2. Photons associated with the correct track are contributing to the Cherenkov peak while photons associated with the wrong track are contributing to the background. The track-photon association algorithm makes more likely to associate tracks and photons when the photon is in the expected angle and as a result the background is not flat, but can be correctly parametrized using a second degree polynomial. The histogram is fitted with a Gaussian and a second degree polynomial and the resolution is given by the Gaussian sigma. In a typical run the value for RICH 1 is 1.63 mrad and for RICH 2 0.68 mrad, very close to the expected design values (see Ref. [3]). It is expected that with more data and better alignment the Cherenkov angle resolution will improve further. 3

1000 900

RICH2 Gas Cherenkov Resolution

Events / ( 0.0001 rad )

Events / ( 0.0002 rad )

RICH1 × Gas 10 Cherenkov Resolution

30000

LHCb Preliminary s = 7 TeV Data

25000

800

LHCb Preliminary s = 7 TeV Data

20000

700 600

15000

500

σΔ θ = 0.001618 ± 0.000002 rad

-0.005

0

σΔ θ = 0.00068 ± 0.00002 rad

10000

400 0.005

Δθ (rad)

-0.002

0

0.002

Δθ (rad)

Fig. 4. Cherenkov angle resolutions for the gas radiators (RICH1 on the left and RICH2 on the right).

3.2. Particle identification performance Particle Identification (PID) is achieved by combining information from all radiators together and treating all the tracks in the event at the same time using a global likelihood algorithm [9]. The RICH detectors operate in a high occupancy environment and the global algorithm provides optimal performance as the Cherenkov rings mostly overlap. All tracks in the event are initially assigned as pions and then the likelihood function is calculated for all the other mass hypotheses (K, p, μ, e). Different cuts in the likelihood difference between mass hypothesis can be used to obtain particle samples with high efficiency or high purity. In order to evaluate the particle identification performance of the RICH system pure collections of particles from the various species are required, identified in LHCb without the use of the RICH detectors. It is possible to obtain these collections from decays that can be identified with high purity using only kinematic information. The V0 decays K s0 → π+ π− and Λ0 → pπ− can be used to provide pure samples of protons and pions. For kaons the decay φ(1020) → K + K − can be used, where a hard cut is applied in the RICH PID of the first kaon to identify the decay and the second kaon is used for particle ID evaluation purposes (“tag and probe” method). The samples of these particles can be seen in Fig. 5. The K s and Λ0 samples are very clean. The φ decay has more background so a special technique called sPlots [10] is used to take the background into account. Using these particle samples it is possible to evaluate the performance of the RICH detectors for particle identification. The result can be seen in Fig. 6, where the performance in terms of efficiency and mis-identification between pions and kaons is shown. Two curves are shown; one for a sample with high efficiency (ΔLL(K − π) > 0) and one for a sample with high purity but somewhat reduced efficiency (ΔLL(K − π) > 5). The performance of the detector has also been evaluated in terms of number of primary vertices (see Fig. 7). LHCb has been taking data under conditions that exceed the original specifications of

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

Fig. 5. Particle collections identified without RICH information which are used for the evaluation of the PID performance of the RICH detectors. The K s and Λ0 collections are very clean. For the φ(1020) the sPlots technique is used to treat the background.

the experiment, however, particle identification from the RICH system remains robust even with multiple primary interactions. 4. Contribution to Physics results In order to demonstrate the power of particle identification from the RICH within a physics analysis two such examples are presented. In the conference report Measurement of direct CP violation in charmless charged two-body B decays at LHCb [11] the authors have selected B decays to two hadrons. Without the use of RICH particle ID the selection is dominated by the B → Kπ decay (Fig. 8). Using RICH particle ID it is possible to isolate the far less common B → ππ and B → KK decays, in good agreement with the the Monte-Carlo simulation shown in Fig.1. A second example of how the RICH information can be used to reject combinatorial background can be seen in Measurement of the relative yields of the decay modes B0 → D− π+ , B0 → D− K + , B0s → D−s π+ , and determination of f s / fd for 7 TeV pp collisions [12]. Fig. 9 shows the invariant mass of the B meson for the selected decays, and in both cases the combinatorial background is almost negligible. 5. Conclusions The RICH detectors are an indispensable part of the LHCb experiment providing particle identification of high accuracy and efficiency. After the alignment and calibration of the system the detectors are operating very close to design performance. The particle identification algorithms have been evaluated with independently selected particle collections and are performing very well, while the majority of the LHCb physics results relies on particle identification provided by the RICH detectors.

617

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

1.4 1.2

Pion Mis-ID Efficiency

Efficiency

618

Δ LL(K - π) > 0

LHCb

Δ LL(K - π) > 5

s = 7 TeV Data

1 0.8

K→K

0.6 0.4 π→K

0.2 0

20

No. Reco. PVs in Event 1 2

10-1

3 4

LHCb Preliminary

10-2

40

60

80

×103 100

Momentum (MeV/c) Fig. 6. Efficiency and mis-identification probability when identifying kaons and pions. Different cuts in the difference between the pion and kaon likelihood values (ΔLL(K − π)) can be applied resulting in particle selection with high efficiency, or very high purity with somewhat decreased efficiency.

s = 7 TeV Data

0.6

0.7

0.8

0.9

1

Kaon ID Efficiency Fig. 7. Efficiency versus mis-identification probability when identifying kaons and pions for different numbers of primary vertices. The performance drops as expected with multiple primary interactions due to the increased number of tracks and photons, however the change is not significant.

References The LHCb collaboration, A.A. Alves Jr. et al., “The LHCb detector at the LHC”, JINST 3 (2008) S08005. The LHCb Collaboration, B. Adeva et al., “Roadmap for selected key measurements of LHCb”, arXiv:0912.4179v2 [hep-ex]. D. L. Perego, The Ring Imaging Cherenkov detectors of the LHCb experiment, these proceedings R. Young, Nucl. Instr. and Meth. A 639 (2011) 94 A. K. Jain, Fundamentals of Digital Image Processing, Prentice-Hall Inc., Englewood Cliffs, New Jersey, US, 1989 M Clemencic J. Phys.: Conf. Ser. 119 (2008) 072010 W. Baldini, et al., LHCb alignment strategy, CERN-LHCB-2006-035, 2006 A. Gorisek at al.,Nucl. Instr. and Meth. A 433 (1999) 408 R. Forty Nucl. Instr. and Meth. A, 433 (1999), p. 257 M. Pivk, Nucl. Instr. and Meth. A 555 (2005) 356-369 Measurement of direct CP violation in charmless charged two-body B decays at LHCb, The LHCb Collaboration, LHCb-CONF2011-011 [12] Measurement of the relative yields of the decay modes B0 → D− π+ , B0 → D− K + , B0s → D−s π+ , and determination of f s / fd for 7 TeV pp collisions, The LHCb Collaboration, LHCb-CONF-2011-013 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

Fig. 8. Event selection from the LHCb conference contribution Measurement of direct CP violation in charmless charged two-body B decays at LHCb [11]. In the top histogram, without RICH information, the selection is dominated by the decay B → Kπ. The far less common decays B → π+ π− and B → K + K − can be successfully selected when the RICH PID information is used. The main components contributing to the model: B0 → Kπ (red), wrong sign B0 → Kπ (dark red), B0 → π+ π− (light blue), B0s → K + K − (dark yellow), B0s → πK (green), combinatorial background (grey) and 3-body partially reconstructed decays (orange).

619

A. Papanestis / Physics Procedia 37 (2012) 613 – 620

Events / ( 8 MeV/c 2 )

620

600 LHCb Preliminary 500

s = 7TeV Lint ~35 pb-1

400 Bd→ D*π Bd→ Dπ Combinatorial Bd→ Dρ

300 200 100 0 5000

5200

5400

5600

5800

Events / ( 16 MeV/c 2 )

Mass (MeV/c2)

100

LHCb Preliminary s = 7TeV

80

-1

Lint ~35 pb

Bd→ D*K Bd→ DK Bd→ Dπ Combinatorial Bd→ Dρ

60 40 20 0 5000

5200

5400

5600

5800

Mass (MeV/c2) Fig. 9. Event selections from the LHCb conference contribution Measurement of direct CP violation in charmless charged two-body B decays at LHCb [12]. The combinatorial background is negligible in both selections mainly due to the RICH PID.