Investigation of Avalanche Photodiodes for EM Calorimeter at LHC. LPNHE, Ecole Polytechnique, F Palaiseau, FRANCE. B. Ille, D.

Size: px

Start display at page:

Download "Investigation of Avalanche Photodiodes for EM Calorimeter at LHC. LPNHE, Ecole Polytechnique, F Palaiseau, FRANCE. B. Ille, D."

Everett Kerry Shelton
5 years ago
Views:

1 CMS TN/9- X-LPNHE/9- October 99 Investigation of Avalanche Photodiodes for EM Calorimeter at LHC A. Karar, Y. Musienko, R. Tanaka, J. C. Vanel LPNHE, Ecole Polytechnique, F-98 Palaiseau, FRANCE B. Ille, D. Si Mohand IPN, Lyon, F-696 Villeurbanne, FRANCE J.P. Bard, J.P. Pansart, J.M. Reymond, J. Tartas DAPNIA Saclay, F-99 Gif-sur-Yvette, FRANCE E. Guschin INR, RU-7 Moscow, RUSSIA Th. Flugel, D. Renker PSI, CH- Villigen, SWITZERLAND J.E. Bateman, S.R. Burge, R. Stephenson RAL, Didcot OX QX, UK P. Cushman, R. Rusack University of Minnesota, Minneapolis MN-, USA S. Reucroft, D. Ruuska Northeastern University, Boston MA-, USA Talk given by Yuri Musienko at \The First Workshop on Electronics for LHC Experiments", Lisbon, - September 99. On leave from INR, Moscow, Russia

2 Investigation of Avalanche Photodiodes for EM Calorimeter at LHC A. Karar, Y. Musienko, R. Tanaka, J. C. Vanel LPNHE, Ecole Polytechnique, F-98 Palaiseau, FRANCE B. Ille, D. Si Mohand IPN, Lyon, F-696 Villeurbanne, FRANCE J.P. Bard, J.P. Pansart, J.M. Reymond, J. Tartas DAPNIA Saclay, F-99 Gif-sur-Yvette, FRANCE E. Guschin INR, RU-7 Moscow, RUSSIA Th. Flugel, D. Renker PSI, CH- Villigen, SWITZERLAND J.E. Bateman, S.R. Burge, R. Stephenson RAL, Didcot OX QX, UK P. Cushman, R. Rusack University of Minnesota, Minneapolis MN-, USA S. Reucroft, D. Ruuska Northeastern University, Boston MA-, USA Abstract Silicon avalanche photodiodes (APD) have been studied extensively as a photodetector candidate for the electromagnetic calorimeter of the CMS detector at LHC. This report presents measurements of the latest generation APDs made by Hamamatsu and EG&G, with particular emphasis on spectral response, excess noise factor, response to charged particles and radiation hardness. The strategy for the future R&D program is also discussed. Introduction The distinctive features of APDs particularly suitable for scintillation detection are: small size, internal gain, insensitivity to the magnetic eld, good matching of their spectral sensitivity with the emission spectrum of the scintillators and low power consumption. Recent developments enabled the fabrication of large-surface APDs (by Advanced Photonix, Hamamatsu, EG&G etc.). However, their application as the photodetector at LHC will require: ) operation in Tesla solenoidal magnetic eld, ) linear dynamic range of, ) high quantum eciency over a wide range of wavelength (- 6 nm), ) good gain stability over the bias voltage and temperature variations, ) negligible \nuclear counter effect" (low response to ionizing particles), 6) small excess noise due to multiplication process, 7) small capacitance and leakage current, and 8) high radiation resistance. The conventional PIN photodiode turned out not On leave from INR Moscow, RUSSIA to be useful due to the non-negligible nuclear counter effect. Therefore, point ) is of particular importance for the CMS electromagnetic calorimeter (ECAL) where the scintillation crystal PbWO is the baseline option []. Three dierent types of APDs have been tested. The two APDs from Hamamatsu (low and high capacitance, type S, mm in diameter) are based on epitaxial growth technology. The third, from EG&G, is a \reverse reach-through" APD (type C66E, mm ) based on ion implantation and diusion technology. The presumed electric eld proles of these APDs are shown in Fig.. The Hamamatsu `low capacitance' APD has a large depletion region in front of multiplication zone. In the case of the Hamamatsu `high capacitance' and s, this region is smaller (? m). In order to decrease the APD capacitance, EG&G introduced a special depleted region behind the gain region. Measurement of The evaluation of all APD parameters (such as spectral response, noise, excess noise factor, response to charged particles etc.) requires correct measurements of its gain. To perform these measurements, dierent signal sources have been used: - continuous light, -7 nm (with the use of a spectrophotometer); - pulsed light (blue or green LED); - gamma sources ( Am, Fe, 7 Co etc.). The simplest way to evaluate the gain of APD is to measure the photocurrent versus bias voltage under continuous light illumination. For this kind of measurement,

3 Hamamatsu S ( 9 pf, φ mm ) Electric Field µm region Depth Number of counts / channel 7.9 kev Hamamatsu Low Capacitance APD 7.6 kev Electric Field region 7 Hamamatsu S ( pf, φ mm ) Depth. kev 6. kev 9. kev ADC channel 6 µm Electric Field Figure : Am spectrum for the Hamamatsu `low capacitance' APD. ( pf, x mm ) region 77 µm Figure : Electric eld prole of the APDs tested. Photocurrent, na H.V. bias, V Depth Figure : Photocurrent versus bias voltage for the \reverse reach-through" type. it is necessary that the light be absorbed before the gain region. If the light is not fully absorbed before the multiplication region, some of the primary photoelectrons will undergo smaller multiplication. Once the APD internal structure is known, one can choose an appropriate wavelength using existing data []. For example, the absorption length at nm is about m. Therefore almost all light is absorbed within m of silicon. The measured photocurrent versus bias for the EG&G APD is shown in Fig.. There is a \plateau" at small biases which also exists for Hamamatsu APDs. Although the APD is not fully depleted with low voltage bias, the rise in photocurrent from V up to biases where the gain starts is very small. This enables us to make a hypothesis that the gain at the \plateau" is close to unity and can be used as a reference. This also means that the quantum eciency of the APD for the light absorbed before the avalanche does not depend on the gain. The value of the gain measured with continuous light at the bias where the APD starts to be fully depleted is used as a reference for the calculation of the gain measured with the LED pulse. The gain measured with the LED pulsed light at higher bias coincides within.% accuracy with the values found using continuous light []. Another way to determine the gain is to expose the APD to -rays (from Am, Fe, 7 Co etc. sources). Some -rays are absorbed in the layer before the avalanche region. A typical spectrum (obtained with the Hamamatsu `low capacitance' APD) is presented in Fig.. From the position of the -peaks in the spectrum, it is then possible to calculate the gain. For gains smaller than -, gains measured using -rays coincide with gains measured with continuous light, with good accuracy. However, for high gains, gains found with -

4 Hamamatsu Low Capacitance APD with Led with Am with Fe Q.E., (=) Crystal luminescence Hamamatsu EG&G Luminescence Arbitrary unit Bias, V 6 Wavelength, nm Figure : versus bias voltage with dierent sources for the Hamamatsu `low capacitance' APD. sources are smaller than those found with continuous light. The higher the gamma energy, the greater the effect (Fig. ). For the, we found an even more complicated eect [, ]. A reasonable explanation of this phenomenon is the saturation eect of the multiplication process due to the \screening eect" caused by high density charge created in the avalanche. The very dense cloud of holes (more than m? ) created in the avalanche locally \screens" the electric eld and decreases the gain. To conclude, coincidence of the gain values in the range of small gains (where APDs are already fully depleted) measured with light and gamma source conrms the hypothesis that the \plateau" of photocurrent versus bias can be used as a reference (gain = ) for calculation of the gain. Thus we can use the simple, fast, accurate and reproducible \continuous light method" for calibration of the gain. Spectral Response The quantum eciency of Hamamatsu and s has been measured at small biases, which correspond to the \plateau" region (Fig. ), using the commercially available spectro-photometer \Graseby Optronix S7" []. A typical light emission spectrum of PbWO crystal is also shown for comparison []. A good match between the spectral sensitivities of APDs and the emission spectrum of the crystal is seen. Values of Q.E. found for the Hamamatsu `low capacitance' APD coincide within % accuracy with those of `high capacitance' type. Figure : Quantum eciency (Q.E.) of APDs (left scale) and crystal luminescence (right scale). Excess Noise Factor One of the most important parameters of APDs is the excess noise factor. The excess noise is due to the statistical nature of the multiplication process, which causes additional uctuation of the measured signal. The energy resolution of an electromagnetic calorimeter can be expressed as follows: E E = p a b c; () E E where the rst, second and third term in right part of the equation are respectively stochastic, noise and constant terms. In the case of small crystal light yield, the main contribution to the stochastic term will be due to photon statistics (we can neglect the intrinsic p resolution of the crystal, which is smaller than %= E(GeV) [6]). For example, if the measured number of primary electrons is 6 per GeV, the stochastic term a will be equal to.%. In the case of APD readout, an additional uctuation of the signal due to the excess noise will be added as: r F a = % N ; () where N is the number of primary photoelectrons per GeV, F is the excess noise factor of the APD at given gain. If the excess noise factor is equal to and N = 6, the stochastic term a will be %. This is why the development of an APD with a small excess noise factor is important for the CMS electromagnetic calorimeter. Standard techniques have been used to measure the excess noise factor []. APDs have been illuminated with blue LED pulsed light. The signal amplitude and its variance have been measured with an ADC. The normalized total variance of the signal is composed of statistical and

5 Exess Noise Factor Hamamatsu Low Capacitance APD + F (green LED) x F (blue LED) Hamamatsu High Capacitance APD noise uctuations: t A F = N + el ; () A with N = Q=M, where N is the number of primary photoelectrons, Q is the charge measured at the input of preamplier, M is the gain of the APD, A is the signal amplitude, t is the r.m.s. of the signal, el is the r.m.s. of the electronic noise, and nally F is the excess noise factor. Using equation (), one can calculate the values of F in terms of measured variances of signal and electronic noise, amplitude of the signal and gain. Fig. 6 shows measured values of the excess noise factor as function of gain for all APDs tested. In all cases, F rises with the gain. For the Hamamatsu `high capacitance' APD, values of F are smaller than for the `low capacitance' type or the. However, to obtain good noise performance with the Hamamatsu `high capacitance' APD, one should operate at higher gain than for the other two types, due to its much higher capacitance. + F (Blue LED) Nuclear Counter Eect Exess Noise Factor Exess Noise Factor + F(Blue LED) Figure 6: Excess noise factor as function of gain measured by LED pulsed light. Charged particles crossing the APD create, due to ionization, approximately 8 e/h pairs per m of silicon (for MIP). As a crystal has a limited longitudinal depth ( X ), the leakage of the electromagnetic or hadronic shower from the rear surface of the crystal delivers some charges in the depleted region of the APD. Charges created before the avalanche region multiply with the same gain coecient as light. On the other hand, charges created in the depleted region located behind the avalanche region (as in the case of the ) undergo much smaller multiplication since the multiplication factor of holes is much smaller than that of electrons. Due to limited photon statistics of PbWO, the signal due to shower leakage into the APD can be non-negligible compared to scintillation signal. In energy spectrum, it manifests itself as a tail in the higher energy part of the Gaussian distribution and causes a perturbation in energy resolution. A matrix of PbWO crystals with APD readout has been exposed to high energy electrons ( GeV) and muons ( GeV) at the CERN SPS-H beam line, in order to study the scintillation and ionization signals. As the sensitive area of the APD is only -6% of the crosssection of the rear surface of the crystal, most muons only cause a scintillation signal in the crystal. However, muons crossing the APD create an additional signal in the depleted region of the photodetector. The scintillation signal from muons is compared with the signal from high energy electrons (or photons) for energy calibration. It is equal to MeV for a GeV muon in X PbWO crystal. Thus, one can express the ionization signal in energy units. The value of this signal decreases slightly for higher gain. For an APD gain approximately equal

6 HAMAMATSU high capa APD Figure 7: Reconstructed energy distribution for GeV electrons with Hamamatsu `high capacitance' and EG&G APDs as photodetector. to, we found 78, and MeV for Hamamatsu `low capacitance', `high capacitance' and s, respectively. For MIP, the ionization loss in silicon is. times smaller than for GeV muon. The energy spectra for GeV electrons are shown in Fig. 7. Hamamatsu `high capacitance' and s are used as photodetectors. For the data taken with the, one can see a small tail in the high energy part of the energy spectrum. It is almost invisible for data taken with the Hamamatsu `high capacitance' APD. For the Hamamatsu `low capacitance' APD, the inuence of the \nuclear counter eect" on energy resolution is quite signicant []. Figure 8: Equivalent noise charge in dependence of neutron uence tested for Hamamatsu `low capacitance' and s. 6 Radiation Hardness At LHC, the ux of neutrons in the barrel ECAL region is expected to be n/cm =year. Such a high ux of neutrons will certainly damage the APD structure. As a consequence, a decrease of the Q.E. and the gain are expected as well as increase of the bulk leakage current due to creation of acceptor- and donor-like levels in the silicon forbidden gap. An irradiation with neutrons has been performed at Ulysse reactor at Saclay [7]. The neutron spectrum of this reactor resembles that expected at LHC with the maximum of energy around MeV. Hamamatsu `low capacitance' and s have been irradiated under bias corresponding to gains between and 7. One month after the last irradiation (total absorbed dose was n/cm ), both diodes had a dark current of the order of : A at room temperature for gain. We have not

7 Quantum Excess Temperature Bias Voltage Response APD Eciency (%) Noise Factor dm dt M (%) dm dv M (%) on MIP () nm (G=) (G=) (G=) (MeV) Hamamatsu l.c. (C d = 9 pf) 6 (8) %. -. Hamamatsu h.c. (C d = pf) 6 (8) % EG&G (C d = pf) 7 (7) % Table : Summary of APD measurements. observed signicant variation of the gain and the Q.E. at xed voltage and temperature. In Fig. 8, the equivalent noise charge, referred to the input of the preamplier, is shown as function of neutron uence. At high uence, both diodes demonstrate about the same noise performance, which is proportional to the gain in the rst order. As the dominating noise source is the bulk leakage current (which is proportional to the damage created by neutrons in depleted region of the APD), it can be reduced by decreasing the operational temperature of the APD and to use ltering with short shaping time. At a temperature of - degree and shaping time equal to ns, the noise of the APD can be decreased by a factor of. to. The Hamamatsu `high capacitance' APD, under bias, has been exposed to an 8 Mrad gamma dose using 6 Co source [8]. An increase of times in leakage current has been observed without signicant degradation of noise performance. 7 Conclusion and Future Plans Three types of latest generation APDs from Hamamatsu and EG&G have been investigated as a potential candidate for the photodetector of the CMS electromagnetic calorimeter. The most important results of the measurement are listed in Table. It was found that at least two APDs, the Hamamatsu `high capacitance' and EG&G, perform close to the CMS requirements. However, CMS in collaboration with Hamamatsu and EG&G continues the work on development of new APDs which will combine the best features of existing prototypes. Specications for the next generation are: - the amplication region should be as close as possible to the surface to minimize the nuclear counter eect (this can also improve radiation resistance due to smaller thickness of the depleted region in front of an avalanche region); - the capacitance of the APD should not exceed pf/cm ; - the dark current should be less than - na/cm at gain ; - the excess noise factor should be less than at gain ; - enhanced photosensitivity in the region of short wavelength (- nm); - the sensitive area should be. cm or more; - the gradient of the gain due to temperature and voltage should not exceed % per degree and less than % gain change per Volt at M=. We also plan to continue radiation hardness and lifetime tests to predict the evolution of APD parameters in the LHC environment. References [] \Compact Muon Solenoid. Technical Proposal", CERN/LHCC 9-8 (99). [] \Handbook of optical constants of solids", vol. Ed. E.D. Palik, Academic Press, 98. [] A. Karar, Y. Musienko, R. Tanaka, J.C. Vanel, \Investigation of Avalanche Photodiodes for EM Calorimeter at LHC", CMS-TN/9-9. [] J.E. Bateman, S.R. Burge and R. Stephenson, \ and noise measurements on two avalanche photodiodes proposed for CMS ECAL", Rutherford Appleton Laboratory, CCL-TR-9-. [] O.V. Buyanov et. al., \A rst electromagnetic calorimeter prototype of PbWO crystals", Nucl. Instr. and Meth. A9 (99) 6. [6] Yu.D. Prokoshkin, A.V. Shtannikov, \Energy resolution calculation of PWO calorimeter, comparison with the beam tests", Nucl. Instr. and Meth. A6 (99) 6. [7] J.P. Bard, J.P. Pansart, J.M. Reymond, J. Tartas, \Results on APD fast neutron irradiations ", to be published. [8] E. Lorenz, S. Natkaniec, D. Renker, B. Schwarz, \Fast readout of plastic and crystal scintillators by avalanche photodiodes", Nucl. Instr. and Meth. A (99) 6. 6

Studies of large dynamic range silicon photomultipliers for the CMS HCAL upgrade

Studies of large dynamic range silicon photomultipliers for the CMS HCAL upgrade Yuri Musienko* FNAL(USA) Arjan Heering University of Notre Dame (USA) For the CMS HCAL group *On leave from INR(Moscow)