Paper submitted to the 2 nd International Conference on Radiation Effects on Semiconductor Materials, Detectors and Devices, held at the Grand Hotel Mediterraneo, Firenze, Italy, March 4-6, 1998, to be published in NIMA Abstract Leakage current of hadron irradiated silicon detectors - material dependence M. Moll, E. Fretwurst, G. Lindström ROSE/CERN-RD48 collaboration II. Institut für Experimentalphysik, Universität Hamburg, Luruper Chaussee 149, D-22761 Hamburg, Germany The leakage current increase of silicon detectors irradiated with fast neutrons was measured in the fluence range from 10 11 to 10 15 cm -2 for a wide range of different starting material. The oxygen concentration in the investigated silicon varied from 9 10 17 cm -3 to below 2 10 14 cm -3 and the carbon concentration from 2 10 16 cm -3 to below 5 10 15 cm -3. Furthermore the resistivity differed from 100 Ωcm to 20 KΩcm for the used n-type and from 400 Ωcm to 2 KΩcm for the p-type silicon. It is found that up to the highest fluence the current related damage parameter alpha is almost independent of the initial resistivity and impurity concentration. After irradiation a universal annealing behavior for all material is observed which unmistakably indicates that the microscopic defects underlying the generation of leakage current are of intrinsic origin. Furthermore it is shown that the parameterization of the annealing behavior at an elevated temperature (here 60 C) provides an excellent tool for comparing the so-called hardness factors of different irradiation sources. As an example the hardness factors for the 24 GeV/c CERN Proton Synchrotron and the TRIGA research reactor in Ljubljana were determined to be 0.51 and 0.76 respectively. 1. Introduction The goal of the ROSE/CERN-RD48 collaboration is to develop more radiation hard silicon detectors that can operate beyond the limit of present-day devices [1]. A key idea for improving the radiation tolerance of silicon is the defect engineering strategy. Since oxygen and carbon act as sinks for primary induced vacancies and interstitials they play a very important role in the defect kinetics. From theoretical calculations it has been predicted that a higher oxygen concentration should have a beneficial effect on the radiation induced changes of the effective doping concentration [2,3]. Together with some experimental evidences for the influence of oxygen on the damage parameters of silicon detectors [4] this gave us the impulse to compare the radiation hardness of silicon containing different oxygen and carbon concentrations. By using Czochralski (Cz) grown silicon ([O] = 9 10 17 cm -3 ), standard float zone (FZ) silicon ([O] 5 10 15 cm -3 ) and FZ silicon with a very low oxygen concentration ([O] < 2 10 14 cm -3 ) we were able to compare material differing in more than 3 orders of magnitude in [O]. While the beneficial influence of high oxygen concentration with respect to the changes in the effective doping concentration for the material used in this work has already been published [5] we present in this paper that there is however no influence of [O] and [C] on the leakage current of hadron irradiated silicon. 3. Leakage current annealing The increase of the leakage current I arises mainly from bulk generation centers introduced by the irradiation. Accordingly, the change I at total depletion is normalized to the sensitive volume V of the device. The variation of current with fluence Φ is expressed in terms of the damage rate α by Corresponding author; e-mail: moll@sesam.desy.de 1
(eq. 1) I ( Φ, T, t) = α( T, t) Φ = R( T) α( TR, t) Φ V where R(T) is the ratio between the leakage current measured at temperature T and a certain reference temperature T R. R(T) is given by [6] as: (eq. 2) R( T) = I( T ) TR = I( T) T E exp 2k B 1 TR. T 2 R g 1 In this work all presented data have been normalized to T R = 20 C using an effective energy gap of E g = 1.12 ev. Most previously reported data on the annealing of the leakage current have been parameterized with a normalized annealing function g(t) consisting of a sum of exponentials [7,8] i i a = (eq. 3) g t) a exp( t / τ ); = 1. ( i It should be mentioned that different parameter sets (, τ i ) have been found in [8] for type inverted and not inverted detectors. The temperature dependence of the annealing process itself has been included into the parameterization in reference [9,10] in a pragmatic way by introducing a temperature dependent scaling factor for the time axis a i (eq. 4) E I θ ( TA ) = exp k B 1 1 TR TA with an activation energy of E I = (1.09±0.14) ev. Taking this temperature dependence into account the current related damage coefficient can be expressed as function of time t at a given annealing temperature T A ( TR, t R ) (eq. 5) α( TR, t) = g( θ ( TA ) t) g( t ) α R where t R is a certain reference time at T R = 20 C after irradiation. From the data of [8,9] it has been deduced that after long term annealing t R constant values α and g are achieved. Therefore, in all previous publications the term α /g was used as the normalizing factor in (eq. 5) [9,10]. In the following we will show that the present results, using an improved experimental method, do not justify this approach any more and that in the contrary α( T R, t) is steadily decreasing with annealing time t. 2. Experimental procedures The device parameters are listed in Table 1 in which the given oxygen and carbon concentrations have been measured by IR absorption and/or SIMS [4,11-14] prior to the processing of diodes. As indicated in some of the cases these values are below the detection limits of the used methods. The broad range of [O] and [C] in the epitaxial p-type material produced by ITME reflects the inhomogeneous depth profile derived from SIMS-measurements [4]. The different manufacturer of the diodes are also given in Table 1. It should also be noted that the thickness of epitaxial diodes cannot be measured as accurately as for the other cases. 2
Crystal Producer crystal Producer diode Guard ring Resistivity [kωcm] [O] [10 16 cm -3 ] [C] [10 16 cm -3 ] α (80min, 60 C) [10-17 A cm -1 ] n-type FZ Wacker a) MPI e) Yes 2.7 < 5 < 0.5 3.99 ± 0.14 n-type FZ Wacker a) ELMA f) Yes 10-20 < 5 < 0.5 4.01 ± 0.04 n-type FZ Wacker a) ITE g) Yes 4.0 < 0.02 < 3 3.87 ± 0.07 n-type FZ Wacker a) ITE g ) Yes 0.42 < 10 < 2 4.02 ± 0.11 n-type FZ Topsil b) Sintef h) Yes 6.6 < 5 < 0.5 4.14 ± 0.06 n-type FZ ITME c) ITE g ) Yes 0.78 17 < 2 3.79 ± 0.08 n-type FZ ITME c) ITE g ) Yes 0.11 < 10 2 3.61 ± 0.11 n-type FZ ITME c) HH-Schottky i) No 0.13 < 10 2 3.93 ± 0.13 n-type Cz Polovodice d) HH-Schottky i) No 0.14 90 0.5 3.94 ± 0.18 p-type EPI ITME c) DIOTEC j) No 0.4 4-20 1-2 4.41 p-type EPI ITME c) DIOTEC j) No 1.6 3-20 1-2 3.92 ± 0.19 p-type EPI ITME c) DIOTEC j) No 3.9 4-60 1-2 4.06 ± 0.40 a) Wacker AG, Burghausen, Germany c) Institute of Electronic Materials Technology, Warsaw, Poland e) MPI-Halbleiterlabor, München, Germany g) Institute of Electron Technology; Warsaw; Poland i) Schottky diodes produced in our laboratory b) Topsil, Frederikssund, Denmark d) Polovodice, Prague, Czech Republic f) ELMA, Moscow, Zelenograd, Russia h) Sintef, Oslo, Norway j) MESA structures produced by DIOTEC, Slovakia Table 1: Silicon material of investigated devices. Note that several process technologies have been used for manufacturing the diodes (MPI, ELMA: implanted diodes; ITE, Sintef: diffused diodes, HH-Schottky: Schottky diodes, DIOTEC: mesa diodes). The samples were exposed to fast neutrons of the d+be source at the PTB Braunschweig/Germany [15]. All neutron fluences given in this paper are normalized to a 1 MeV equivalent neutron beam using the appropriate hardness factors between 1.45 and 1.50 depending on the individual distance of the samples from the target [16,17]. In the time interval between the end of irradiation and the beginning of the first measurement the samples were stored in liquid nitrogen to avoid any annealing process. For the investigation of the annealing process itself an isothermal heat treatment at 60 C was chosen to accelerate the process with respect to room temperature. The first point used for evaluation of the annealing function was measured after 5 min at 60 C which already corresponds to an equivalent annealing time of about 15 h at room temperature. Hence the self annealing during the extended irradiations (up to 9 h to reach 1 10 15 cm -2 ) does not play any significant role. Between the isothermal heating steps the I/V characteristics were measured at room temperature. In order to extract reproducible and reliable damage parameters from standard I/V characteristics the guard ring of the diodes was connected to ground when ever possible (see Table 1). As an example in Figure 1. I/V characteristics of three type inverted detectors are shown. It can be seen that measurements without a connected guard ring lead to drastically wrong leakage current values due to the indeterminate active volume of the diode. Therefore the values used for our analysis were always the saturation currents measured at high voltages with connected guard rings. It has to be emphasized that most of the deviations between the results presented in this paper and the ones of previous publications result from the different techniques with respect to the use of the guard ring. 3
100 I[µA] 10 1 MPI ITE ELMA 0.1 0.1 1 10 100 voltage[v] Fig. 1. Examples of I/V characteristics for three type inverted diodes from different manufacturers. The filled symbols indicate the measurements with and the open ones those without connected guard ring. 4. Experimental results 4.1 Annealing at 60 C For numerous detectors fabricated from different material as listed in Table 1 the isothermal annealing of the leakage current at T A = 60 C has been measured. The extracted α values as function of the cumulated annealing time are plotted in Fig. 2 for devices irradiated in a fluence range between 2.5 10 11 cm -2 and 1.7 10 14 cm -2. The solid line is a best fit to the presented data and is described by (eq. 6) α 60 C, t) = α exp( t / τ ) + { α β ln( t / )}. ( 1 1 0 t0 The corresponding parameters achieved from a fit to all measured data are summarized in Table 2. α 1 10-17 A/cm τ 1 min α 0 10-17 A/cm β 10-18 A/cm t 0 min 1.01±0.38 93±24 5.03±0.09 3.34±0.26 1 Table 2: Parameters of the current annealing (eq. 6) at T A = 60 C. t 0 was set to 1 min. For comparison the dashed line in Fig.2 represents the model prediction given by (eq. 5) with α = 2.86 10-17 Acm -1 [9] and the parameters of the annealing functions g(t) for not inverted detectors given in Ref. [8]. A time scaling factor of θ = 177 (eq.4) was used as calculated with an activation energy of E I = 1.09 ev [9]. Although some of the high resistivity n-type samples are type inverted and some are not, in contradiction to Ref.[8] no differences in the annealing behavior can be seen. Furthermore, the annealing curve follows well a logarithmic time dependence in the long term and no final constant value 4
of α is observed. This differences are most probably due to the fact that the diodes used in [8] had no guard ring. α [ 10-17 A/cm ] 8 7 6 5 4 3 2 1 0 simulation fit 1.7 10 14, DIOTEC 2.9 10 11, ELMA 2.5 10 12, ELMA 2.3 10 13, ELMA 2.5 10 11, MPI 1.3 10 13, MPI 5.8 10 12, ITE 5 10 1 5 10 2 5 10 3 5 10 4 time at 60 o C [ min ] Fig. 2. α values as function of cumulated annealing time at 60 C for some diodes. The leakage current was measured at room temperature and normalized to 20 C (eq. 2). The legend gives the neutron fluence and the manufacturer of the diode (see Table 1). Note that some of the samples are type inverted and some not. Although the discrepancy between the old model (Ref. [8,9]) and our new parameterization are obvious (see Fig. 2) we are not able to present a new temperature dependent parameterization due to the lack of measurements at temperatures other than 60 C. While for the exponential term in (eq.6) the time scaling by (eq.4) works well the logarithmic term in (eq. 6) can not be scaled in the same way. Therefore, further isothermal annealing experiments are necessary and under way to determine the activation energy for this part of the annealing function. Meanwhile first measurements show that the activation energy is higher than 1.1 ev. 4.2 Material dependence In order to compare the different material the α values measured after 80 min at 60 C are plotted in Fig. 3 as function of the fluence and are summarized for the different materials in Table 1. It is very striking that the current damage rate α does not depend on the material properties of the silicon, i.e. the doping and impurity concentration. For fluences higher than 1 10 14 cm -2 α seems to decrease slightly. A possible explanation could be that for such high fluences the concentration of the defects (disordered regions) that are responsible for the leakage current are in the order of the impurity concentrations and, therefore, are influenced by the migrating interstitials and vacancies produced by the irradiation. This, however, has to be investigated in further experiments. From the wide range of investigated materials it can be concluded that after irradiation with energetic hadrons mainly intrinsic defects composed of vacancies and interstitials are responsible for the increase of the generation current but not impurity related defects. 5
5.0 4.5 α [10-17 A/cm] 4.0 3.5 3.0 2.5 2.0 n-type FZ - 10 to 20 KΩcm n-type FZ - 7 KΩcm n-type FZ - 4 KΩcm n-type FZ - 3 KΩcm p-type EPI - 2 and 4 KΩcm n-type FZ - 780 Ωcm n-type FZ - 420 Ωcm n-type FZ - 130 Ωcm n-type FZ - 110 Ωcm n-type CZ - 140 Ωcm p-type EPI - 380 Ωcm 10 11 10 12 10 13 10 14 10 15 Φ eq [cm -2 ] Fig. 3. α values measured after 80 min at 60 C for several kind of materials (see Table 1). The solid line represents the mean value of (3.99 ± 0.03) 10-17 A/cm in the fluence range of up to 2 10 14 cm -2 while the dashed lines indicate the one sigma level of 0.24 10-17 A/cm. Structures without guard ring are indicated by filled symbols. 4.3 Determination of hardness factors In addition to the neutron irradiations at the PTB some samples (ELMA, see Table 1) have been irradiated with 24 GeV/c protons at the CERN PS [4] and some (ITE 800Ωcm, see Table 1) with neutrons at the TRIGA research reactor in Ljubljana [18]. Due to the transport of the samples from CERN or Ljubljana to our laboratory after irradiation these samples could not be stored in liquid nitrogen and have been exposed for about three days to room temperature before starting the isothermal heat treatment at 60 C. The α values measured after 80 and 2880 min at 60 C are shown in Figure 4. With an acceleration factor of at least 180 with respect to room temperature the measurement at 2880 min represents an annealing time of at least 1 year at room temperature. This means that no influence of the annealing period at room temperature before the heat treatment is expected to be seen in the leakage current any more. Therefore the heating procedure together with the parameterization of the annealing function at 60 C gives a powerful tool to compare hardness factors of different sources independent of irradiation and annealing time at room temperature before the heat treatment. Compared to α measured on identical samples after neutron irradiations at the PTB the hardness factor for the 24GeV/c protons was determined to be 0.51 ± 0.01 and the one of the TRIGA reactor to 0.76 ± 0.04. The given error is of statistical nature only. For systematic errors in determining hardness factors see Ref. [16]. The measured hardness factor for the protons is in good agreement with the experimental value of 0.58 ± 0.03 given in Ref.[10] and fits to the theoretical value predicted by Huhtinen (about 0.5) [19] while there is no accordance with the older value of 0.93 given by van Ginneken [20]. For the TRIGA research reactor a hardness factor of 0.88± 0.05 has currently been calculated from the measured spectrum in reference [21]. The discrepancy to our experimental hardness factor clarifies the need for a 6
standardized method for hardness factor determination of different sources (like the one presented in this work) in order to get a basis for comparing radiation damage experiments. α [10-17 A/cm] 3.5 3 2.5 2 1.5 Ljubljana reactor: 80 min at 60 o C 2880 min at 60 o C CERN PS: 80 min at 60 o C 2880 min at 60 o C 1 10 11 10 12 10 13 10 14 Φ [cm -2 ] Fig. 4. α (not normalized to 1 MeV neutrons) vs. fluence after 80 and 2880 min annealing at 60 C. The lines indicate the mean values (CERN PS protons: 2.01±0.03 and 1.20±0.02 10-17 A/cm; Ljubljana reactor neutrons: 2.91±0.13 and 1.70±0.08 10-17 A/cm) and the one sigma levels (CERN PS: ±0.05 and ±0.07 10-17 A/cm; Ljubljana: ±0.36 and ±0.22 10-17 A/cm). Note that different types of diodes have been used at CERN and Ljubljana (ELMA and ITE-800Ωcm, see Table 1.). 5. Conclusions We have shown that the leakage current damage rate α for hadron irradiated silicon detectors is independent of the fluence in the range from 10 11 to 10 14 cm -2 with only a slight decrease for even higher fluences. Furthermore, α is also independent of the silicon material. The used material varied in the oxygen concentration from 9 10 17 cm -3 to below 2 10 14 cm -3 and in the carbon concentration from 1 10 16 cm -3 to below 5 10 15 cm -3. The resistivity differed from 100 Ωcm to 20 KΩcm for the used n-type and from 400 Ωcm to 2 KΩcm for the p-type silicon. The universal annealing behavior for all this material gives us clear evidence that the microscopic defects underlying the generation of leakage current are of intrinsic origin i.e. they are composed of vacancies and interstitials and are not impurity related. The annealing behavior at 60 C was parameterized as a sum of an exponential and a logarithmic term. While the exponential term was in rather good agreement with data found in literature we could show that, in contrast to former publications, there is no saturation value α for the damage rate α. Finally it was shown that the measurement of α during an isothermal heat treatment at 60 C together with our annealing parameterization provides an excellent tool for comparing hardness factors of different sources. As an example the hardness factor for 24 GeV/c protons was determined to be 0.51 ± 0.01 and the one of the TRIGA research reactor in Ljubljana to be 0.76 ± 0.04. 7
Acknowledgements We would like to thank G. Casse, M. Glaser, F. Lemeilleur and A. Ruzin for including our samples into their irradiation program at the CERN PS and the TRIGA research reactor in Ljubljana. Furthermore we thank R.Böttger and H.J.Brede for providing the irradiation facility at the Physikalisch-Technische Bundesanstalt Braunschweig and P.Buhmann and U.Pein for measuring hundreds of C/V and I/V characteristics. References [1] RD48 STATUS REPORT, CERN/LHCC 97-39, Status Report/RD48 (1997). [2] B.C. MacEvoy et. al., NIMA 374 (1996) 12. [3] B.C. MacEvoy, PhD thesis, Imperial College, London, November 1996. [4] B. Dezillie, PhD thesis, University Joseph Fourier - Grenoble 1, September 1997 (IR and SIMS measurements). [5] E.Fretwurst et al. "A comprehensive model for the effective doping concentration of irradiated silicon detectors" presented at "The Third International (Hiroshima) Symposium on the Development and Application of Semiconductor Tracking Detectors" at Melbourne 9-12th December,1997, to be published in NIMA. [6] S.M. Sze, Physics of Semiconductor Devices, 2nd ed. (Wiley, 1984). [7] F.Lemeilleur et al., Nucl. Instr. and Meth. A360 (1995) 438. [8] R.Wunstorf, Ph.D. thesis, Universität Hamburg, see also DESY FH1K-92-01 (1992). [9] A.Chilingarov et al., Nucl.Instr. and Meth. A360 (1995) 432 and references therein. [10] H.Feick, PhD thesis, Universität Hamburg, see also DESY F35D-97-08 (1997). [11] EVANS EUROPA, Uxbridge, UK (SIMS measurements). [12] P. Clauws, University of Gent, Belgium (IR measurements). [13] ITME, Institute of Electronic Materials Technology, Warsaw, Poland (IR measurements). [14] B.G. Svensson, Royal Institute of Technology, Stockholm, Sweden, (SIMS measurements). [15] H.J.Brede et al., Nucl.Instr. and Meth. A274 (1989) 332. [16] A.Vasilescu, Technical Note ROSE TN97/3, 1997. [17] H.J.Brede and A.Vasilescu, private communication. [18] Jozef Stefan Institute and Department of Physics, University of Ljubljana, SI-1000 Ljubljana, Slovenia. [19] M. Huhtinen and P. Aarino, HU-SEFT R 1993-02 (1993). [20] A. van Ginneken, Non ionizing energy deposition in silicon for radiation damage studies, Fermi Nat. Accelertor Lab. Report FN-522 (1989). [21] D.Zontar et al., "Time Development and Flux Dependence of Neutron-Irradiation Induced Defects in Silicon Pad Detectors", presented at the "2 nd International Conference on Radiation Effects on Semiconductors Materials, Detectors and Devices" to be published in NIM A. 8