Gallium arsenide pixel detectors

Gallium arsenide pixel detectors

NUCLEAR INSTRUMENTS 8 METHODS IN PHYSICS RESEARCH ELSEYIER Nuclear Instruments and Methods in Physics Research A 410 (1998) 6-l SectIon A I G...

416KB Sizes 2 Downloads 68 Views




and Methods

in Physics


A 410 (1998) 6-l

SectIon A


Gallium arsenide pixel detectors R. Batesa, M. Campbellb, E. Cantatoreb, S. D’Auria”, C. DaVii”, C. de1 Papa”**, E.M. Heijneb, P. Middelkampb, V. O’Shea”, C. Raine”, I. Ropotarb, L. Scharfetterb, K. Smith”, W. Snoeysb

RD8 Collaboration Aachen,










Sheffield, Tomsk.

Udine, Vilnius

Abstract GaAs detectors can be fabricated with bidimensional single-sided electrode segmentation. They have been successfully bonded using flip-chip technology to the Omega-3 silicon read-out chip. We present here the design features of the GaAs pixel detectors and results from a test performed at the CERN SpS with a 120GeV K- beam. The detection efficiency was 99.2% with a nominal threshold of 5000e-. I<’ 1998 Elsevier Science B.V. All rights reserved.

1. Introduction A single-sided, bidimensional position sensitive detector with high spatial resolution is an extremely useful tool in high-energy physics, in situations where the density of charged particles is high. The high level of segmentation greatly reduces the occupancy and is invaluable in pattern recognition. as a pixelated detector gives a true unambiguous point in 3-D. A small volume element also improves the radiation hardness because the larger reverse bias leakage current is shared among a large number of amplification channels. Silicon 2-D single-sided sensors are presently operational in HEP experiments [l&3]. These detectors can be



E-mail: delpapa(

016%9002!98/$19.00 F 1998 Elsevier Science B.V. All rights reserved PII: SO168-9002(98)00094-l

connected with flip-chip bonding to the same front-end electronics developed for silicon sensors without any change in the process. GaAs detectors have been commercially fabricated with a footprint compatible with the Omega-3/LHCl chip, developed at CERN. Here we present results from a 120 GeV/c rt test beam at the H6 secondary beam of the CERN SpS.

2. Sensor design The detectors characterised in this paper have been fabricated by a GaAs foundry’ with a 4-mask

1 Alenia S.p.A.. direzione Roma, Italy.


Via Tiburtina

km 12.4

R. Bares et al. /Nucl. Instr. and Meth. in Phys. Rex A 410 (1998) 6-11


Fig. I. A pictorial

view of the detector,


the GaAs sensor. the bump bonds

photolithographic process on semi-insulating LEC-grown GaAs substrates. Surface barrier metal contacts are patterned with a lift-off technology. A silicon nitride layer is used as passivation. To make the process compatible with solder bump bonding a metal stopping layer must be deposited on the bonding apertures, to prevent the solder from reaching the semiconductor or penetrating underneath the passivation. An electroplated gold layer is grown on the bonding pads to facilitate the adhesion: it will form a eutectic alloy with the solder, which in this process is deposited on the read-out chip only. The fabrication of GaAs pixel detectors is compatible with the fabrication process of strip detectors, so both devices can be accomodated in a single test run. The back contact was also patterned, implanted with Sif ions and thermal activated to allow overdepletion. This feature makes these detectors innovative with respect to those described in previous reports [4]. The flip-chip assembly was made by a commercial firm,’ using a single-sided solder bump technique. A sketch of the assembled device is shown in


and the electronics



Fig. 1. The back contact was not protected with passivation. Some scratches due to handling during the flip-chip assembly are possible causes of the limited operating voltage we have found in some of the detectors. To avoid this a protective layer will be deposited during the device fabrication in the future. The working detectors allowed an operating voltage up to 500 V, without breakdown, where the full thickness is active for detection. This value is larger than for GaAs detectors with no ion implantation and is probably due to the thermal processing. The pixel geometry has the footprint of the ‘Omega-3/LHCl’ read-out electronics chip developed by the RD19 collaboration at CERN. The pitch is 50 x 500um2, the metalization is 40 x 490 urn’. The array consists of 16 x 128 pixels, covering an area of 51.2mm’, surrounded by a 200um wide guard ring. The substrate thickness was 200um and the saw cut was made at about 200 urn from the guard-ring.

3. Test beam: Set-up and results ‘GEC-Marconi tester, Northants

Materials Technology NN12 8EQ, England.

Ltd., Caswell,


The set-up, shown in Fig. 2, consists of three layers of single-chip pixel detectors. The two outer


Silicon Pixeldetector

Fig. 2. A pictorial view of the test beam detector are described in the text.




set-up with a pixel tclescopeand

phsrjc scintillators

for trigger. The srnsor dimensions


Q 2CO & ; 000 ;


Fig. 3. A beam protile measured with the GaAs detector AL-IOC-IO at Vhi4, = 500V with a I70GeWc dimensions are determined by the scintillation counters providing the event trigger.


beam at the SpS. The

R. Bates et al. ,lNucl. Instr. and Meth. in Phlr. Rex ‘4 410 (199H) 6-/l

ones were silicon sensors and were our beam telescope. The trigger was generated by the coincidence of 4 photomultiplier signals. The trigger area was defined by two scintillating fibres, each 1 mm in diameter, overlapping for about 2.5 mm in length. The Omega-3/LHCl read-out electronic chip [S] consists of a preamplifier, a comparator and a delay line for each channel. The comparator threshold setting is common to all the pixels. The delay line output is enabled while the external strobe signal is active. In this test the duration of the strobe signal was loons, so that an individual fine delay adjustment on every pixel was not needed. A GaAs detector ‘hit’ was considered valid if it matched within 200um in the ‘vertical’ direction and 500um in the horizontal direction a track found in the two reference silicon pixel detectors. The noisy pixels, for the smallest threshold value, were recorded with beam-off conditions and a random trigger run and masked during data collection. The alignment was based on the correlation plots between hits in the three detectors. This method was possible because of the extremely low divergence of the beam. The 2-D histogram of the number of valid hits per pixel versus the pixel position gives the convolution of the beam profile with the


scintillator local efficiency and is shown in Fig. 3. The rounded shape is due to the cylindrical shape of the fibres. The uniformity of the contacts was excellent, with only 3 pixels which were probably not connected. The number of noisy pixels which were turned off with a mask was 5. A systematic threshold variation was found throughout the chip [6]. This makes it difficult to give a precise number for the mean threshold in the beam region, but the threshold found in an earlier absolute calibration was used [S]. The systematic variation has been corrected in a more recent version of the front-end chip [6]. The track detection efficiency is defined as the number of valid hits in the GaAs detector divided by the number of reconstructed tracks in the two silicon pixel planes. The variation of the track detection efficiency versus threshold is shown in Fig. 4, for three different values of detector bias. At low bias the signal is still too low to reach full detection efficiency, but at 500 V a plateau with full efficiency is evident. By comparison with the same curve for a 300 urn thick silicon sensor (not shown) the GaAs curve has the same plateau value, but it then decreases faster. As the calibration curve strongly depends on the settings of the front-end



s .-5


: 80 0


70 60 50 40 30 20


10 0





20000 1



Fig. 4. Track detection eficiency averaged over the trigger area of the GaAs pixel detector integrated discriminators. The maximum value is 99.2% with a 5000e- threshold.




versus the common



of the


Fig. 5. Hit multiplicity pions


in the 200 pm thick GaAs detector

chip and on the geometrical position of the beam spot it would be misleading to measure the collected charge from Fig. 4, which shows an efficiency of 99.2% with a nominal threshold of 5000 eP and the presence of a plateau. The charge sharing was also measured: 19.4% of the tracks, which were perpendicular to the sensor, gave signals on two adjacent pixels, as shown by the multiplicity distribution in Fig. 5. This value is in good agreement with the geometrical aspect ratio (metal/gap = 40/10) and is also in agreement with the simulation of GaAs microstrip detectors [7].

4. Conclusions

and discussion

The GaAs detector technology maturity has been established via a large production of charged particle detectors with a satisfactory yield. Bump bonding of GaAs detectors can be performed with essentially the same yield as for silicon devices, provided that an appropriate choice of metals is

from single track events due to normally



made for the bonding pads. The nitride passivation can stand the thermal stress during fabrication and the assemblies which are made are both mechanically and electrically stable. The yield and the reliability can be improved by depositing a protective layer on the back contact, to avoid scratches and allow overdepletion. Ion implanted back contacts give excellent detectors at the expense of a higher operating voltage, of about 2.5 V/urn. The pixel detectors show a good detection efficiency plateau. higher than 97% with a comparator threshold of up to 8000e-. The feature of a faster signal was not used as the shaping time was imposed by the front-end VLSI amplifier. Further tests will be made with a shorter strobe gate. but we expect no substantial difference from a 300um thick silicon sensor. It has been shown recently [8,9] that semi-insulating GaAs is not as radiation resistant as was indicated by earlier measurements [lo], especially with regard to the damage due to charged hadrons. The technology described above is, in any case. now available for both microstrip and pixel

R. Bates et al. /Nucl. Instr. mrd h&h.

detectors. There is hope that enhanced radiation hardness will be provided by some new kind of GaAs material. For the present, GaAs devices can be used at basically the same cost as silicon if a faster signal or a higher atomic number is required.

in Phys. Rex .A 4/O (1998) 6-11

111 F. Antinori VI 131

Acknowledgements We gratefully acknowledge the assistance of the many people who have been involved in the device preparation and testing. In particular, we thank A. Cetronio and C. Lanzieri from Alenia S.p.A., G. Humpston from GEC-M, F. Cossey-Puget and J.P. Avondo for their technical support, E. Chesi, H. Beker and A. Mazzoni for help in setting up the DAQ. Sensor fabrication was made possible by funds from I.N.F.N. We are also grateful for the support of this work from PPARC through a CASE postgraduate studentship sponsored by the Rutherford-Appleton Laboratory.


c41 c51 [61


C81 c91 Cl01

et al.. Nucl. lnstr. and Meth. A 360 (1995) 91. E.H.M. Heijne et al., IEEE Trans. Nucl. Sci. 42 (4) (1995) 413. K.H. Becks et al.. Nucl. Instr. and Meth. A 386 (1997) 11. S. D’Auria et al., Nucl. Ins&. and Meth. A 386 (1997) 177. E.H.M. Heijne et al., Nucl. Instr. and Meth. A 383 (I) (1996) 55. E. Cantatore et al., Uniformity response of the LHCl large area pixel detector system, Frontier detectors for frontier Physics, La Biodola, 25531 May, 1997, to appear in Nucl. Instr. and Meth. A. R. Bates et al., in: P.G. Pelfer et al., (Eds.). Proc. 3rd Int. workshop on GaAs detectors and related compounds, San Miniato, Italy. 21-24 March 1995, World Scientific (1996) 92. R.L. Bates et al.. Nucl. Instr. and Meth. A 395 (1997) 54. F. Tenbush et al., Nucl. Instr. and Meth. A 388 (1997) 38. S.P. Beaumont et al., Nucl. Instr. and Meth. A 322 (1992) 472.