ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 604 (2009) 202–206

Contents lists available at ScienceDirect

Nuclear Instruments and Methods inPhysics Research A

0168-90

doi:10.1

E-m1 O

journal homepage: www.elsevier.com/locate/nima

Hybrid Photon Detectors for the LHCb RICH: Performance andoperational experience

Laurence Carson 1

Department of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ, UK

a r t i c l e i n f o

Available online 14 February 2009

Keywords:

LHCb

RICH

Pixel HPD

02/$ - see front matter & 2009 Elsevier B.V. A

016/j.nima.2009.01.198

ail address: l.carson@physics.gla.ac.uk

n behalf of the LHCb RICH collaboration

a b s t r a c t

Pion/kaon discrimination in the LHCb experiment will be provided by two Ring Imaging Cherenkov

(RICH) counters. These use arrays of 484 Hybrid Photon Detectors (HPDs) to detect the Cherenkov

photons emitted by charged particles traversing the RICH. The results from comprehensive quality

assurance tests on the 550 HPDs manufactured for LHCb are described. Leakage currents, dead channel

probabilities, dark count rates and ion feedback rates are reported. Furthermore, results from two

specialised measurements carried out on a sample of HPDs to determine the efficiencies of the HPD

pixel chip and photocathode, respectively, are described. Finally, an overview is given of the in situ

performance of the HPDs which are now installed in the RICH.

& 2009 Elsevier B.V. All rights reserved.

1. Introduction

The LHCb experiment [1] is an experiment that will run at theLarge Hadron Collider (LHC) at CERN, Geneva. When data takingbegins in 2008, LHCb will commence the study of rare processesin the B-hadron system with the aim of discovering New Physics.Accurate and reliable Particle Identification (PID) will be crucial tothe success of the LHCb physics program. To this end the LHCbdetector includes two Ring Imaging Cherenkov (RICH) detectorswhich will identify pions, kaons and protons over a widemomentum range of 1–100 GeV/c.

Both RICH detectors make use of arrays of pixel Hybrid PhotonDetectors (HPDs) [2] to detect the Cherenkov photons which areemitted by charged particles as they traverse the RICH. The RICHesuse 484 HPDs between them, 196 being in RICH1 and 288 being inRICH2. The HPDs must meet stringent performance requirementsif they are to provide PID capability over the lifetime of the LHC.This paper reports the results of comprehensive quality assurancetests on all 550 HPDs which were manufactured for LHCb, andalso gives an overview of the in situ performance of the HPDswhich are now installed in the RICH.

2. The pixel HPD

The pixel HPD combines the advantages of vacuum and silicontechnology by encapsulating a silicon sensor inside a vacuum-sealed tube. Fig. 1 shows a schematic diagram of an HPD.

ll rights reserved.

The principle of operation of an HPD is as follows: an incomingCherenkov photon reaches the quartz window of the HPD, andstrikes the layer of S20-type multialkali photocathode materialwhich is deposited on the vacuum side of the window. This liberatesa photoelectron, which is then accelerated and focused onto theanode by a 20 kV electric field. The anode consists of a pixellatedsilicon diode which is bump-bonded to a binary (pixel) readoutchip. The binary readout chip has 8192 pixels. During LHCb runningthese are ORed in groups of 8 to make 1024 superpixels.

The HPD manufacturing process involved a number of differentinternational companies, with the final stages carried out byPhotonis-DEP.2 The quality assurance tests were carried out byLHCb between October 2005 and July 2007 using dedicatedPhoton Detector Test Facilities (PDTFs) at the University ofGlasgow and the University of Edinburgh.

3. Quality assurance tests

All 550 HPDs which were manufactured for LHCb have nowbeen tested at the PDTFs. Table 1 shows some of the specificationsfor the performance of the manufactured HPDs, compared tothe actual average performance of the HPDs from tests carried outat the PDTFs. The collective performance of the HPDs in the areasmost crucial for their physics performance is further illustrated inFigs. 2–5. Note that the whole sample of 550 HPDs contained only12 HPDs (i.e. 2.2% of the sample) which had a performance issueserious enough to disqualify them from use in the RICH.

2 Photonis Netherlands B.V., Dwazziewegen 2, P.O. Box 60, NL-9300 AB Roden,

Netherlands. Formerly Delft Electronic Products.

ARTICLE IN PRESS

Table 1Selected specifications and actual HPD performance.

Property Specification Average performance

Working pixels 95% min 99.8%

Photoelectron detection threshold 1500 e� max 1064 e�

Photoelectron detection noise 100 e� typical 145 e�

Photoelectron detection efficiency 85% typical 88%

Leakage current at 80 V reverse bias 1mA typical 1:49mA

Quantum efficiency at 270 nm 20% min 30.9%

Dark count rate 5 kHz=cm2 typical 2:54 kHz=cm2

Ion feedback probability 1% max 0.03%

Fig. 2. Silicon sensor leakage current (nA) at 80 V reverse bias.

Fig. 3. Number of dead pixels (out of 8192).

Fig. 1. Schematic of the pixel hybrid photon detector.

L. Carson / Nuclear Instruments and Methods in Physics Research A 604 (2009) 202–206 203

It can be seen that the leakage current is normally less than1mA, which is the specified typical leakage current. A few HPDsexhibit high (45mA) leakage current, but this is not a problem aslong as the amount of voltage lost is small enough such that thesensor is still fully depleted when 80 V reverse bias is applied.Only one HPD failed to fully deplete due to high leakage current,and it was rejected. The dead pixel probability is always less thanthe specified maximum of 5%, which corresponds to 410 deadpixels. At the time of measurement, no HPD exceeds the specifiedmaximum of 1% ion feedback probability. Some HPDs do exceedthe specified maximum dark count level of 5 kHz=cm2, but theseHPDs were not rejected as the dark count level is still 1000 timeslower than the level where it would adversely affect the physicsperformance of the RICH. In addition, the higher dark count rateis correlated with the improved quantum efficiency (QE) of theHPDs.

4. Specialised tests

The tests described in Section 3 were carried out on all 550HPDs. In addition there are two further tests which, due to theirtime consuming nature, were only carried out on a subset of HPDs.One of these tests measures the detection efficiency of the siliconsensor/binary readout chip ensemble, and the other measures theQE of the photocathode.

Fig. 4. Ion feedback probability (%). This measures the quality of the vacuum

inside the HPD by measuring the number of ions created by photoelectrons as they

pass through the vacuum.

4.1. Backpulse measurements

The number of photoelectrons seen by the binary readout chipwill be lower than the number arriving at the backplane of the

ARTICLE IN PRESS

Fig. 5. Dark count levels (kHz=cm2). This is the noise level seen when the high

voltage is on but no light is incident on the HPD. Fig. 6. Fit to backpulse spectrum. The y-axis shows the number of counts per

channel, accumulated over a period of around one hour. The 2nd, 3rd and 4th

photoelectron peaks can be clearly seen, with the other peaks appearing as

shoulders due to the high noise level.

Fig. 7. ZSi results for HPD H630005, using a binary readout time of 50 ns. This HPD

has ZSi ¼ ð92� 2Þ%. Note that ZSi is independent of the amount of light input

(x-axis).

L. Carson / Nuclear Instruments and Methods in Physics Research A 604 (2009) 202–206204

silicon sensor because some photoelectrons do not depositenough charge within the readout time of the binary readoutchip to reach the detection threshold. This is due to charge-sharing and backscattering effects, and defines the photoelectrondetection efficiency ZSi of the silicon sensor/binary readout chipensemble. A specialised test to measure ZSi (the ‘‘backpulsemeasurement’’) was carried out on two HPDs. This measurementinvolves shining light from an LED onto the quartz window, andcounting the number of photoelectrons this produces by twodifferent methods. Firstly, the number of photoelectrons strikingthe backplane is estimated by measuring the amount of chargedeposited there. This is done by tapping into the backplanedirectly and then passing the signal through a preamplifier and abuffer amplifier. Over many events, a histogram of the charge canbe built up, which should show peaks at multiples of the chargedeposited by one photoelectron, which is around 5000 electrons.The number of photoelectrons at the backplane is then measuredby fitting a custom-defined function [3] to the charge spectrum.One of the parameters in the fit is the (Poisson mean) numberof photoelectrons per event. A typical spectrum (black dots) andfit (green line) is shown in Fig. 6.

Secondly, the standard PDTF Labview software is used to countthe number of hits registered by the binary readout chip. This isthen divided by the number of photoelectrons seen at thebackplane to give ZSi. The measured value of ZSi depends on theamount of time (known as the binary readout time) allowed tothe binary readout chip to integrate the signal from the siliconsensor. Fig. 7 shows the ZSi values obtained for one of the twomeasured HPDs, tube H630005, using a binary readout time of50 ns. The average value across the two HPDs is ZSi ¼ ð94� 2Þ%.The errors on each ZSi value are estimated by consideringthe change in the light output level of the LED during a singlemeasurement, and by examining the stability of the fit to thebackpulse spectra. Measurements were also carried out ontube H630005 using a binary readout time of 25 ns, as this iswhat will be used during LHC running. These yield ZSi ¼ ð88� 2Þ%,which as expected is lower than the 50 ns value.

4.2. QE measurements

The QE of the photocathode was measured in a standard wayby biasing the photocathode relative to the anode then illuminat-ing the photocathode with monochromatic light and comparingthe current at the photocathode to the current across a

photodiode of known QE. The QE measurement was carried outfor every HPD by Photonis-DEP, and for a subset of HPDs by LHCbat the PDTFs. In general excellent agreement was found betweenPhotonis-DEP measurements and LHCb measurements (see Fig. 8).

During the production process, Photonis-DEP achieved con-sistent improvement in the QE of manufactured HPDs, with theQE value at 270 nm rising from around 26% in early batches ofHPDs to around 32% for later batches. The fact that the QE of allHPDs is higher than expected (see Table 1) will directly improvethe performance of the RICH.

5. Commissioning and monitoring

All 484 HPDs have been mounted on columns and installed inthe RICHes in the LHCb experimental area. A thorough programmeof in situ commissioning and monitoring is well underway. Fig. 9

ARTICLE IN PRESS

A_posXY_10_6_L0name_55OFF!

L. Carson / Nuclear Instruments and Methods in Physics Research A 604 (2009) 202–206 205

shows the whole photodetector plane of RICH2 under illuminationby continuous laser light.

One of the monitoring systems being used is the magneticdistortion monitoring system, or MDMS. This system measuresthe deflection of the path of the photoelectrons due to the residualfield (maximum value of around 2.4 mT) from the LHCb magnet.This deflection must be measured so that it can be corrected forwhen finding Cherenkov rings.

H703012 Quantum Efficiency0.400

0.350

0.300

0.250

0.200

0.150

0.100

0.050

0.000

QE

, ηηq,

(no

units

)

0 100 200 300 400 500 600 700 800 900 1000Wavelength λ / nm

DEP ResultsPDTF Results

Fig. 8. Comparison of QE measurements made by Photonis-DEP and PDTF on a

single HPD. Excellent agreement between the two measurements is found across

different wavelengths.

HP

D R

ow

0000

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

C0

C1

C2

C3

C4

C5

C6

C7

C8

HPD Co

Pixel Hit Map - R3M Events of Run 24716, taken on 1

HV = 18kV, CW Laser, IFB Clustersize Threshold = 5, R

Fig. 9. Illumination of RICH2 HPD pla

To measure the deflection the HPD plane is illuminated with aspecific pattern of light, and the deformation of the pattern ismonitored as the LHCb magnet is powered on and off. The patternon one HPD in RICH2 is shown in Fig. 10.

A1

A2

A3

A4

A5

A6

A7

A8

A0

lumn

1

10

102

103

104

ich2, All Hits5 Jun 2008, ALICE mode Analysisun 25611 referenced for Dark Count Compensation

ne using continuous laser light.

Pixel Column0 5 10 15 20 25 30

Pix

el R

ow

0

5

10

15

20

25

30

10

102

103

104

Fig. 10. Illumination of an HPD in RICH2 by the MDMS system.

ARTICLE IN PRESS

L. Carson / Nuclear Instruments and Methods in Physics Research A 604 (2009) 202–206206

6. Conclusions

All HPDs required by the LHCb RICH detectors, includingspares, have now been produced and thoroughly tested. Only 12out of 550 HPDs fail to meet the performance requirements. Theremaining HPDs meet all specifications and exceed them incertain key areas, for example number of working channels, darkcount rate and QE. The manufacture and testing of the wholesample of HPDs required for LHCb has proceeded according to

plan. Commissioning of the HPDs in the RICH detector is wellunderway, and the LHCb RICH collaboration are awaiting the firstLHCb data with optimism.

References

[1] A. Augusto Alves Jr., The LHCb Collaboration, et al., JINST 3 (2008) S08005.[2] M. Moritz, et al., IEEE Trans. Nucl. Sci. NS-51 (3) (2004) 1060.[3] T. Tabarelli de Fatis, Nucl. Instr. and Meth. A 385 (1997) 366.