Dense lead glass shower counters at high energies

Nuclear Instruments and Methods 215 (1983) 93-101 North-Holland Publishing Company

DENSE LEAD GLASS SHOWER COUNTERS AT HIGH ENERGIES Shuji ORITO

Deparment of Physics, Universitv of Tokyo, 7-3-1 Hongo, Bunhy o-ku, Tokyo. Japan 113

Tornio KOBAYASHI and Hiroshi TAKEDA

LICEPP, Faculty of Science, Universitv of Tokvo, 7-3-1 Hongo, Bunkyo-km Tokyo, Japan 113

Received 13 December 1982 Systematic tests on dense lead glass shower counters were carried out by using electron and hadron beams in the momentum range between 0.3 GeV/c and 90 GeV/c. An array of Nikon DF6c blocks (density 5.2 gem - ;. radiation length 1 .7 cm) provided an excellent shower-energy resolution : (aE /E = 3.63/CE +0 .04%, E in GeV) approaching 0.4% at 90 GeV. A uniformity better than t0 .3% over the surface, shower position resolutions of 7 to 2 mm, and pion rejection ratios of 10 2 to 10 3 were obtained for beam momenta of 1 to 40 GeV/c. Effects of materials in front of the counters were also studied. 1. Introduction Lead glass Cherenkov counters have been widely used for the detection of photons, electrons and positrons in high energy experiments . So far SF-2 and SF-5 (or equivalent lead glasses) were mainly used [1,21 because of their superior transparencies, and it has been generally believed that denser lead glasses result in worse energy resolutions due to their poorer transparencies and higher refractive indices, which make optical coupling to photomultipliers less efficient. On the other hand, as the energies of the particles increase Table I Properties of various lead glasses. PbO content (weight %) Radiation length " Xo (cm) Radiation length XMC, (cm) r:t::.al e::'sroy (Mev) -~ Density (g/ cm`) Refractive index (n d ) Nuclear collision length(cm)

F-2

SF-2

SF-5

DF6c/SF6

45

51

55

71

3 .06

2 .76

2 .54

2 .94

2 .59

2 .23

18 .4

17 .5

15 .8

1 .70 1 .53 13 .8

3.86

4 .08

5 .20

1 .620

1 .648

1 .673

1 .805

21 .8

21 .4

2. Experimental apparatus 1.1 . The lead glass Cherenkox counter

3.60

22.7

to several tens of GeV, longer lead glass blocks must be used to contain all the showers in order to maintain the good energy resolution . Eventually_ the lengths required become prohibitive. An alternative possibility is the use of denser glasses with shorter radiation lengths. Such glasses are expected to provide better electron and pion separations since they provide less probability of nuclear interactions for a fixed total radiation thickness (see table 1) . We ha%e tested counters made of DF6c glass. by Nikon. at energies from 0 .3 to QO GeV;-c using beams at K1-K . DESY and CERN . Test results are reported in this paper on the energy resolution, uniformity, position resolution, e/ir separation and the effects of materials in front of lead glass counters .

17 .8

"' Conventional radiation length [41.

h ' Radiation length according to Meswl and Crawford [9].

0167-5087,'83/0000--0000/$03 .00 , 1983 North-Holland

We of the lead glass counters is sketched in fig. 1 . The lead glass block was made of Nikon DF6c [31. This glass has a density of 5.2 gem _, a refractive index n a = 1 .805, and a radiation length [41 of 1.70 cm . The properties of this glass are compared with other glasses in table 1 . Each glass block is wedge-shaped with front and rear faces of 8.9 x 10.3 cm= and 10.3 x 10.3 cm =. respectively, and a length of 37 cm corresponding to 21 .8 radiation lengths. All faces were polished optically. The internal light transmission of each block was measured over the thickness of 37 cm . All the lead glass blocks used here show internal transmission inside the

S. Orito et al. / Dense lead glass shoAwr counters

8 so

Fig. 1. Sketch of a lead glass counter.

0

GM Wove length of light, X(nm)

hatched area of fig. 2. To study the dependence of the energy resolution on the length of the blocks, 30 cm and 41 cm long blocks were also used. Each lead glass block was wrapped with aluminized mylar for optical isolation and viewed by a Hamamatsu R1618 photomultiplier (5]. This photomultiplier has a 3" bialkah photocathode with a maximum quantum efficiency of 30%. The linearity of the photomultiplier % as measured to be better than 1% up to 40 mA. The entrance glass of the photomultiplier was Pyrex with a refractive index nd = 1 .489. The photomultiplier was glued to ,he rear face of the glass block with an optical epoxy glue Cemedine-1565 (6] (n d = 1 .56) . A It-metal magnetic shield surrounded the photomutplier . We arranged these lead glass counters in a 3 x 3 matrix array (or a 3 x 4 matrix for the measurement of uniformity and position resolution). The blocks were packed tightly together with gaps less than 0.5 mm between them . The whole array was then wrapped in a black vinyl sheet for light sealing.

Fig. 2. Internal transmittance of DF6c blocks, directly measured over 37 cm thickness. from 8 to 90 GeV/c) . The properties of the beam lines are summarized in table 2. For all the cases the beam was defined by a coincidence of scintillation counters immediately in front of the lead glass. For the CERN beam lines the separation of electrons and pions was obtained below 20 GeV/c by means of two threshold gas Cherenkov counters. Muons were discriminated from pions by using a hadron calorimeter [7] consisting of 16 layers of 5 cm thick iron and 5 mm thick plastic scintillator plates, placed behind the lead glass counters. At the KEK beam, particles were unambiguously identified by means of two threshold gas Cherenkov counters and time of flight measurements. The DESY beam provides pure positrons. Signals from the photomulipliers (13 ns rise time and 30 ns fwhm) were sent to integrating ADCs (Lecroy 2282A [8]) operated with a gate width of 800 ns . The integrated pulse height information was transferred to an HP2100 on-line computer together with information from the gas Cherenkov counters and the hadron calorimeter . In order to equalize the gains of photomultipliers prior to the measurements, the electron beam was di-

'.? . Test beams and data taking

The measurements were carried out using four test beams. KEK -~2 (pions and electrons from 1 to 3 lie% , ( ). DESY beam-26 (positrons from 0.3 to 6 CERN-SPS H3 and H5 (hadrons and electrons Table 2 Description of test beams.

Particle momentum (GeV/c) Momentum spread ap/P(%)

Beam spread or %-idth o¬ defusing counter (mm fv,hm) Material after last bending magnet (rad. length)

CERN-SPS (H3) e-'17

6-90

-

CERN-SPS (1-15) e-,Or 10-40

0.1

0.4

2

1r)

0.16

0.10

DESY Beam 26 e+

KEK 7r2 e t.ir t

0.3-6

1-2

1

1

10

10

0.07

0.15

S. Orito et al. / Dense lead glass shower counters

rected at the center of each counter in turn, and the high voltage was adjusted so that the gains of all counters were set equal to within 1% of each other. Further correction was made in the off-line analysis to an accuracy of ±0.2% . Unless otherwise stated, the outputs from counters were then added for further analysis . 3. Test results 3.1 . Energy resolution Three arrays of 3 x 3 blocks, with glass thicknesses of 30, 37 and 41 cm respectively, were used to measure the energy resolution for electrons. Fig. 3 shows a typical pulse-height distribution for electron and hadron beams of 60 GeV/c. A sharp electron peak is seen, well separated from the hadrons. The pulse-height distribution was first corrected for a slight nonlinearity of up to 3%, measured for each counter thickness by the dependence of the electron peak position on the beam energy. The energy resolution is obtained by fitting the electron pulse-height distribution with an asymmetric Gaussian function : P)2 y= A exp[- (x/2A2], where

=A, forx>p, and A=A-

forx

95

to minimize chi-squared . The resulting resolution is represented by of/E(4 + + A _)/2 p or equivalently by the AE(fwhm)/E = 2.36oE/E. The results in term of fwhm and a E are displayed in figs . 5 and 6 as functions of the beam energy for three different counter thicknesses. As seen in these figures, the counter array of 41 cm thickness exhibits excellent energy resolutions, having 8.6% fwhm (3.6% a) at 1 GeV and 0.99 % fwhm (0 .42% a) at 90 GeV. The 30 cm thick counters show a poorer energy resolution compared to the longer glasses especially at the highest energies . The energy dependence of the resolution can be expressed by a function : aE/E = al CE + b. A leastsquares fit provides : aE/E = (3 .42 ± 0.08)/V + (0 .45 ± 0.03)%, for 30 cm thick counter; = (3 .61 t 0.05) /FE + (0 .11 ± 0 .02)`x. for 37 cm thick counter ;

_ (3 .63 ± 0.05)/F + (0 .04 ± 0.02)%

for 41 cm thick counter; with E in units of GeV. A large constant term b for the 30 cm thick counters is probably due to shower leakage from the back surface. Fig. 4 shows a comparison of the pulse-height distributions of 60 GeV electrons for three different counter thicknesses, together with the fitted

The four free parameters A, p, A, and A _ were adjusted 400 400-

60 GeV e

0)

30 an thick

N

y 4000)

Ó

~6 200 _ .a E

Z

0400-

c)

41 cm thick

200-

Pulseheight (Chonn els) Fig. 3 . Integrated pulse-height spectrum of 60 GeV/c electrons and hadrons measured with 3 x 3 array of 37 cm thick DF6c counters.

0.95

o NorTlized "eheight

1.05

Fig . 4 . Pulse-height spectra of 60 GeV electrons in the lead glass array of thickness a) 30 cm. b) 37 cm and c) 41 cm . The curves are the results of fitting with an asymmetric Gau-ian

S . Orito et al. / Dense lead glass shower counters

Electron energy. E (GOV Fig. 5 . Energy resolution of lead glass counter arrays of various thicknesses as a function of electron energy .

asymmetric Gaussian curves . Since different photomultipliers were used for different glasses. no absolute comparison of the pulse heights was possible and the distributions in this figure have an arbitrary horizontal scale. The effect of the shower leakage is clearly seen for

30 cm thick counters as the wide tail in the distribution at the lower side of the peak . Asymmetric distributions observed for 37 cm and 41 cm thick counters (figs. 4b and c) indicate the importance of shower leakage at high energies even for these thicknesses .

s x

e

DF6c . Sum of 9 counters 0.5

0.1 Ot

e

x --- 300 t

e

" ---370 t

0.5

1 5 Electron energy, E(GOV)

0

10

50

100

Fig. 6 . Energy resolution of lead glass counter arrays of various thicknesses as a function of electron energy .

S. Orito et al. / Dense lead glass

97

shower counters

3.2. Uniforrni~v The position dependence of the electron peak pulseheight was measured using 3 and 30 GeV beams with beam widths (fwhm) of 10 and 2 mm respectively. The incident beams were perpendicular to the surface of the counter array of 3 x 4 blocks with 37 cm thickness, at various positions between the centers of the two central blocks. Results at 30 GeV are shown in fig. 7. As seen in fig. 7a, the sum of the counter signals is constant within the measurement accuracy of f 0.3% except for points very near the gap. Fig. 7b shows the energy sharing between the central two blocks . Fig. 7c shows that the energy resolution is also constant except for the point very near the 0.5 mm gap between the blocks . Similar results for 3 GeV positrons are shown in fig. 8. The dependence on incident angle was also measured for 40 GeV electrons. As seen in fig. 9, the dependence is smooth and the deviation amounts to a maximum of 3% at 15°. This can be easily corrected to an accuracy

(%)

F

(%)

100 r counter-1

50 F

3

a)

center of counter-1

center of counter-2

gap

b)

e

6r

e

w 4 L. .. .

-50

Incident beam ,position

5,_

TT

Fig. 8. Uniformity for 3 Gel' electron (beam Nidth 10 mm fwhm) ; a) position dependence of the summed pulse height . b~ energy sharing between the neighbouring central ttio hl .xks . position dependence of the energc resolution .

of 0.1 % if the incident angle of the particle is known to 2° . 3.3. Position resolution The shower-energy sharing among the blocks provides a method of determining the impact point of electrons and photons to an accuracy much better than the block size . Using the data described in the previous section, the accuracy of the position determination was investigated in the following way. First a weighted average ( .r) of the shower pl, Nitioil was calculated by :

0L (%)5,

C) = a [ ` . . . ."" . . . .. . .1 . .- . . e, e

-50

íJ

1

0

1- .~ ....._.

,. ..._

...

e

_

50 (mm)

Incident beam position Fig. 7. Uniformity for 30 GeV electron (beam width 2 mm fwhm); a) position dependence of summed pulse height, h) energy sharing between the neighbouring central two blocks, c) position dependence of the energy resolution .

= Y_ E"x/' y_ k'," . where x, is the central position of the i th counter and E, represents the energy deposited in the i th counter. The parameter a was experimentally chosen to be 0.4 in order to minimize the discrepancy .x = J.i - x h i, where

S Orito et al. / Dense lead glass shower counters

DF6c, Sum of 4 Counters 40 GeV

t 0

a"

to' 20' Incident beam angle, ® (deg )

30r

40"

Fig. 9. Incident-angle dependence of the summed pulse height for 40 GeV electrons. Counter thickness 37 cm .

xb is the actual beam position. Fig. l0a shows the relation between x and xb for 30 GeV electrons incident

normal to the entrance surface. For each electron shov er, the shower position (x) is then determined from z by

using the calibration curve (solid line ., fio %,a), and the distribution of the deviation x - xr , represents the accuracy of the position determination. The resulting position resolution (ax ) is shown in fig.

50-

X

40-

lob as a function of the beam position after corrections

3C -

for the effect of

finite beam width (2

mm

fwhm).

Resolutions of t 1.2 mm and t 3.2 mm were obtained at the border and at the center of the blocks respec-

tively . Fig. 11 shows the position resolution curve ob-

10

20

30 Xb

40

50

30

40

50

f

60 (mm)

tained by adding the .% - xb distribution at each point assuming a uniform distribution of incident particles

(mrn) 4 Qx 3 ~~t

0

10

20

Xb

60 (mm)

Fig. 10. Shower position determination for 30 GeV electrons: (a) relation between the weighted average (z) and the beam position (zb), (b) position resolution ox as a function of the beam position.

X-Xb (CM ) Fig. I I. Position resolution curve for uniform illumination by 30 GeV electrons.

S. Orito et al. / Dense lead glass shower counters (mm) 10

6x

x 5

e

0

0'

X - .- 3 GeV o --- 30 GeV " --- 40 GeV - 3 GeV,

'

10.

20'

9crn Al in front

3f

9 ( deg,)

Fig. 12. Incident angle dependence of the avecagg, position resolution for 30 Ge%' electrons. over the surface; an accuracy a.,,= ± 2.3 mm can be expected for this case. Similar measurements at 3 and 40 GeV were made for various inclinations of the beam. The resulting óx are summarized in fig. 12 . .í'.4. Electron and pion separation The electron can be identified by comparing the shower energy to its momentum. Hadrons interacting in the counter could, however, simulate electrons. The probabilities of charged pions simulating an electron (R,,,,) were measured[ at l, 2, 10, 20 and 40 GeV/c using the array of 3 x 3 blocks. Figs . 13a and b display shows-r energy distributions in the lead glass counters for c1carly identified charged pions of 2 and 20 GeV/c- .

Fig. 14. Pion rejection ratio vs. electron efficiency . The deposited shower energy was normalized to the peak position of the electron shower energy. From these figures and the shower energy distribution from electrons, R,1 , and the electron efficiencies can be determined as functions of the shower-energy cut. Plotted in fig. 14 are curves of R,, as functions of electron efficiency for various beam energies . Fig. 15 shows R,, , corresponding to an electron efficiency of 95% . as a function of the beam energy. R n,: e scales as 1,,'v'E implying an energy-independent shape of the normalized shower energy distributions for pions near the end point.

2GeVn

Shower energy /E e

0

b

0.2

.4 0

Q6

0 .8

Shower energy/E e

Fig. 13 . Shower energy distributions (normalized to the electron peak position) for (a) 2 GeV pions and (b) 20

1.0 GeV

pions.

S. Orito ,1 a;. / Dense lead glass shou-er counters

100

4

nesses in front of the lead glass counters. The results on energy resolution are summarized in fig. 16 . The effects of the material diminish rapidly as the energy increases. The effects on the position resolution were measured at 3 GeV for normal incidence. As seen in fig. 12, Qt was increased from 6.7 to 7.8 mm by a 9 cm thick aluminum plate. The effects of material on the pion rejection were measured with 8 and 13 .5 cm thick aluminum plates, and are summarized in fig. 15 . A loss of a factor 2-3 in the pion rejection ratio is observed. It should be noted that the effects of material on energy resolution and pion rejection can be minimized by placing presampling detectors between the material and the lead glasses.

f c

0

u m

f

4. Conclusion

" --- no At o --x ---

T0j

1

with 8 cm At with 13.5 cm At

X

5

10

i

50

100

Beam Energy (GeV ) Fig. 15 . Pion rejection ratio for electron efficiency of 95%.

3-5. Effects of materials in front

In some applications, the presence of material in front of the lead glass counters is unavoidable . The effects of materials on the energy resolution, position re-boluuon. and electron-pion identification were investigated by placing aluminum plates of various thick-

Cherenkov shower counters made of dense DF6c glass provide excellent energy resolutions in a wide energy range from 0.3 to 90 GeV/c . With a block length of 24 radiation lengths, a resolution of 0.99% fwhm (0.425f a) can be obtained for 90 GeV electrons. From the energy sharing among 10 .3 cm wide counters, the impact position of the shower can be determined to an accuracy of f2.3 mm and f 6.7 mm, averaged over the surface, at 40 and 3 GeV, respectively. The sum of counter signals shows a uniformity better than ±0 .3% over the surface. Pion rejection ratios of 10 2 to 10', for 95% electron efficiency, were measured in the beam momentum range 1-40 GeV/c. We wish to thank colleagues at the University of Tokyo, KEK, DESY and CERN for their support and assistance . We are particularly indebted to Messrs . J. Kanzaki and S. Odaka who contributed in data taking and analysis of 1 and 2 GeV data. Help from Prof . Y. Totsuka, Drs . A. Sato, M. Minowa, Messrs . T. Mashimo, K. Kawagoe and T. Kajita is greatly appreciated . Thanks are also due to Drs. H. Wenninger and H. Siebert for their collaboration at CERN . We are indebted to Messrs. T. Ichimura and M. Kariya of Nikon for their efforts in providing excellent glass. Finally sincere thanks go to Prof. M. Koshiba for support and encouragement. References [ 11

Fig. 16 . Effect of materials on the ener :;y resolution .

Among others, M. Holder ct al ., Nucl . Instr. and Meth . 108 (1973) 541; J.S . Beale et al . . Nucl . lnsts. and Meth. 117 (1974) 501 ; J.S. Appel et al ., Nucl . Insir. and Meth . 127 (1975) 495; A.S . Carroll et al ., Nucl . Instr. and Meth . 179 (1981) 229; 121 A glass equivalent to DF6c was tested below 1 GeV by Y. Yoshimura et al, Nucl . Instr. and Meth . 126 (1975) 541.

S. Onto et al. / Dense lead glass shower counters [31 Nikon, Nippon Kogaku K .K . Asamizodai 1773, Sagamihara, Kanagawa, Japan . This glass has a similar composition to SF6N by Ohara and SF6 by Schott . [4] H. Rossi, High energy particles (Prentice Hall, Englewood Cliffs, N .Y ., 1952). (5] Hamamatsu TV co., LTD., 1126 Ichino-cho, Hamamatsu. Japan . [6] Cemedine Inc., Shinagawa, Tokyo, Japan .

It)1

[71 Designed by W . Kienzle and 0 . Runolfsson . We thank them for letting us use it . [81 LeCroy Research systems Corp., Spring Valley . N .Y. 10977. 10977, USA, CAMAC 2280 system with 48-channel ADC 2282A . [9] H. Messel and D .F. Crawford, Electron-photon shower distribution function tables (Pergamon . Oxford, 1970).