Comparison and implications of charge collection measurements in silicon and InGaAs irradiated by energetic protons and neutrons

Please download to get full document.

View again

of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Comparison and implications of charge collection measurements in silicon and InGaAs irradiated by energetic protons and neutrons
Document Share
Document Transcript
  IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 42 NO. 6 DECEMBER 1995 1815 Comparison and Implications of Charge Collection Measurements in Silicon and InGaAs Irradiated by Energetic Protons and Neutrons E. Normand, D. L. Oberg, J. L. Wert & P.P. Majewski, Boeing Defense & Space Group Seattle WA 98124 2499 G. A. Woffinden, Amdahl Corp. S. Satoh and K. Sasaki, Fujitsu Laboratory Ltd and Fujitsu Ltd. M. G. Tverskoy & V.V. Miroshkin, Petersburg Nuclear Physics Institute N. Goleminov, Moscow Engineering and Physics Institute S. A. Wender and A. Gavron, Los Alamos National Laboratory Abstract A variety of charge collection measurements by energetic protons and neutrons have been measured and compared. These include deposition in: small silicon junctions, large volume American and Russian silicon surface barrier detectors, and InGaAs photodiodes. INTRODUCTION A limited number of charge collection measurements using high energy neutrons and protons have been reported [l-41. Since these measurements can provide insight into the mechanisms for single event upset and latchup by neutrons and protons in microelectronic devices, it is worthwhile compiling and comparing a diverse set of such measurements. Three different types of charge collection measurements will be discussed a) high energy neutrons in small silicon subvolumes, b) high energy neutrons and protons in large silicon volumes, namely surface barrier detectors (SBDs) and c) high energy neutron in InGaAs photodiodes. Most of the measurements are new. These include the measurements in the small silicon devices by Fujitsu and in the InGaAs photodiodes. The Russian and American SBD results were previously published [ 1-41. However, additional information was obtained relating to the test conditions of the two Russian measurements that enabled the Russian data to be transformed into a different format. This allowed their energy deposition measurements to be directly compared against those obtained from roughly comparable American tests. All the high energy neutron measurements were made at the Weapons Neutron Research WNR) facility at the Los Alamos National Laboratory. For the silicon SBD measurements, these will be compared against energy deposition curves measured using several different high energy proton beams, including two in Russia. Thus, these measurements represent energy depositions at four different proton facilities: the Harvard Cyclotron (154 MeV), TRIUMF (200 MeV) , he Petersburg Nuclear Physics Institute PNPI, 1000 MeV) and the Joint Institute for Nuclear Research (JINR) at Dubna, Russia (3650 MeV). The JINR tests were conducted by Moscow Engineering & Physics Institute (MEPhI) personnel. The silicon SBDs, both those used by Boeing and Clemson University, as well as those in Russia, appear to be quite similar. The other devices are more unique. The small silicon subvolumes were test devices specially developed by Fujitsu Ltd. and Fujitsu Laboratories Ltd., and were tested by personnel from Amdahl Corporation and the two Fujitsu companies. These devices contained two test structures having the same overall area, one comprised of a single structure and the other of 36,000 small junctions. The InGaAs photodiodes are produced by Epitaxx Corp. Three different sized photodiodes were used, the two larger diameter diodes to increase the charge deposition, and the smaller diode to be typical of those incorporated into fiber optic data bus subsystems. SILICON SMALL OLUMES Fujitsu Laboratories and Fujitsu Laboratories Ltd. fabricated chips containing the charge collection test structures. There were two test structures, one a single large structure, 120pm x 1200 pm, and the other an array of 36,000 small structures (each with active area of 2 pm x 2 pm). As shown in Figure 1, the overall active area of each was identical, 1.44 E5 pm2, and each was comprised of the same N+ diffusion (depth of 4 pm) structure in a P+ well over a P-substrate. The array of 36,000 small parallel junctions also had a O:l majodminor axis, i.e., were arranged in a 600 x 60 array. Each square-shaped active area was separated from one another by pm of the p well that served as an isolation barrier. Eight such chips were simultaneously exposed to the high intensity neutron be m of WNR in beamline 30L. Figure 2  shows the energy spectrum of the WNR neutrons. Charge was collected in both structures at 3 and 6 volt bias when the beam was at normal incidence. With the structures at 3V, charge was also collected at incident angles of 0,30,60,90 and 180 degrees. The measured integral charge collection spectrum from the combined array of small structures is shown in Figure 3 in Burst Generation Rate (BGR) units, [i.e. (number counts depositing > E)/(volume x fluence)]. Figure 3 shows that for 0018-9499/95 04.00 0 1995 IEEE  1816 Edep I6 MeV, the variation in the charge collected With the angle of incidence is relatively small (< factor of 2). For energy deposition of >6 MeV, this energy deposition variation with angle of incidence is much more pronounced, and for Edep > 10 MeV, the variation with angle is very large. Slngle large volume (1200 um x 120 pm> / 600 x 60 4pm2 Junctlons IC 4.8 r Array of 36000 small volumes I1 pwell T --+I -1pm passivation si 2 b n+ diffusion I Figure 1 Layout of Chip with Silicon Test Structures, Single Large Volume and Array of Small Subvolumes deposition at 10 MeV represents a small fraction of the total deposition, and is based on a very small number of counts. Figure 4  contains a similar comparison of the measured energy deposition in the single large volume of active silicon. Again, the energy deposition curves are quite similar for Edep -5 MeV, and exhibit very large differences for Edep > 8 MeV. The test devices acquired fluences that ranged between 3- 6E10 n/cm2 during each run (combination of applied voltage and device orientation to the beam). Thus a large number of depositions were recorded during each run, roughly 1E4, spread over a very fine deposition structure of approximately 1000 channels (0.01-15 MeV, 0.0156 MeV/channel), although only the lowest 400 channels received many counts. For the single large structure data of Fig. 4, > 99.5% of these counts were for energy depositions of < 6 MeV. Thus for Edep > 6 MeV statistical uncertainties make it difficult to discem from Fig. 3 whether there is an actual trend in the data regarding the variation of deposition rate with angle of incidence. The BGR format, in which the number of counts is divided by the fluence, normalizes out the total number of counts at each angle, but it doesn't eliminate the poor tatistics of low counts for high energies. lh-IO Figure 2 Ehergy Spectrum of the WNR Neutron Beam (30L) Used for Charge Collection Measurements To assess the statistical uncertainties, a closer evaluation of the actual number of counts in the single large structure used in constructing Fig. 4 is tabulated in Table 1. To make the comparison more consistent between the runs, the number of counts in each ru has been normalized to 1E4 counts (similar to normalizing by fluence). Statistical uncertainties for each normalized number of counts is taken as dcounts. As seen in Table 1, for Edep > 4 MeV and 3V operation, the Table 1 Energy Deposition Counts for Large Edep by Number of Counts Total Cnts > 0.02 Actual Cnts > 4MeV Normlzd Cnts > 4MeV Actual Cnts > 6MeV Norm'lzd Cnts AMeV Actual Cnts > 8MeV lNom'lzd Cnts >8MeV IActuaI Cnts > l0MeV Nom'lzd Cts >loMeV Angle of Incidence  1817 normalized number of counts varied from 149 12 to 200 14 over the five different angles of incidence, indicating fair uniformity over all angles of incidence.. However for Edep > 8 MeV, the normalized number of counts was 18f 4.2 at 90 incidence, but varied between 6-8+ 2.7 for the other four angles of incidence. This same trend of much larger deposition at 90 is even more apparent in the data for Edep > 10 MeV, where at 90 incidence there were 5f 2.2 counts, recorded but none at the other angles of incidence. significantly more high energy deposition than at normal or near-normal angles of incidence. It appears that this is due to deposition from the secondary protons and alpha particles. For grazing angles, these light secondary particles, generally emitted in the forward direction, will be facing much longer paths (minimum of 120 pm) over which to deposit appreciable energy compared to the normal incidence case of -4 pm path length (approximate diffusion depth). This same angular dependence effect (peaking of high energy deposition at 90 ) was also seen in charge collection measurements with Even accounting for the poor statistics, it appears that the 300 MeV protons [ 141. net F -oris striking at 900 (grazing incidence) produce '€ s 5 ti Q m W A c .- 8 d 1 E-1 3 IE-14 1 E-1 5 1 E-1 6 IE-17 1E-18 2 4 6 8 10 12 14 16 Deposited Energy, MeV lure 3 Energy Deposition Curves for Array of Small Silicon Subvolumes in WNR Beam at Various Angles of Incidence 1E-13 1 E-14 1E-15 1E-16 1E-17 1 E-1 8 0 2 4 6 8 10 12 Deposited Energy, MeV Figure 4 Energy Deposition Curves for Single Large Volume in WNR Neutron Beam at Various Angles of Incidence  1818 In Figs. 3 and 4 we also compare calculated values of the BGR deposition by high energy neutrons obtained from the Los Alamos version of the HETC code, LAHET [8]. Unfortunately, LAHET calculations are available for only a limited number of models [9], and none is fully applicable to the test structure geometry and deposition conditions. The closest calculation is for that of a 4x4~4 m3 volume of silicon that only accounts for the energy deposited by the recoils, and a similar calculation based on an infinite medium of silicon. The WNR neutron spectrum is very similar to that of the atmospheric neutrons, having the property that about 1/3 of the neutrons are in each of the three energy ranges: 1- 10 MeV, 10-100 MeV and 100-700 MeV. The calculated BGR energy deposition curves for the WNF? spectrum that are in Figs 2  and 3 are based on averaging the BGR functions for neutrons of three different energies, 8 [lo], 50 and 200 MeV 191, each being representative of one of the three aforementioned broad energy ranges with a nearly equal number of neutrons. Overall, energy deposition by the recoils in the 4x4~4 m3 volume is much more representative of deposition in the test structures than are the recoils in the infinite medium calculation, although for Edep <5 MeV the deposition is relatively similar for the two geometries. For Edep>S MeV, energy deposition in the infinite medium far exceeds that measured in the test structures as seen in Figs. 2  and 3. Out to about Edep -10 MeV, the calculated energy deposition in the 4x4~4 m3 volume is in relatively good agreement with the measured curves, but for Edep>lO MeV, the calculated deposition falls off far too rapidly. This is due to the energy deposition from the light reaction products, i.e., protons and alpha particles, which isn't included in the LAHET calculations. The role of the protons is further discussed below in order to explain other aspects of the measured deposition. In Fig. 5 we directly compare the measured energy deposition in the array of small volumes with that in the single large volume. From Fig. 5 it is clear that for Edep > 5 MeV the deposition is greater in the array than in the smgle volume . There appear to be at least two factors that account for this: a) the effective junction is actually larger in the array due to the additional sidewall junctions between the N+ subvolumes and the surrounding P-well, allowing for additional charge collection, and b) enhanced deposition due to the additional energy degradation in the secondary protons. Energy deposited by protons and alpha particles, which have long ranges compared to the ranges of the neutron-induced recoils (< 8 pm), plays a very important role. The increased energy deposition in the array is partially due to the protons and alpha particles and their energy degradation through the extra silicon isolating each small subvolume. The cumulative effect of the extra silicon surrounding each subvolume is to degrade the energy of the protons further than in the single large volume. Because the dE dx of the protons and alpha particles increases as the energy decreases (for E > 1 MeV), the net result is that the energy deposition is increased. This can be more clearly shown by example, taking a neutron interaction resulting in the production of an energetic proton, (in this case 15 MeV), that is assumed to occur at the center of the volume of active silicon. For a simple 90 scattering, the longest path length through the entire volume is along the major axis; for the single volume it is 600 pm, but for the may, it is 900 pm (cumulatively 600 pm of active silicon and 300 pm of isolating p-silicon). In the single large volume, the energy deposition over the last 2 pm of active silicon is 0.015 MeV, and over the last 10 pm of active silicon, it is 0.074 MeV. By comparison, in the array of small subvolumes the energy deposition over the last 2 pm of active Figure 5 Comparison of Charge Collection in Large and Small Test Structures in WNR Beam at Normal Incidence silicon is 0.018 MeV, and over the last 10 pm of active silicon, it is 0.090 MeV. Thus near the outer edge of the active volumes, approximately 20% more energy is deposited by a 15 MeV proton in the active portion of the array of subvolumes compared to that in the single large volume due to the increased dE/dx of the lower energy protons. This 20% increase is consistent with the increased energy deposition overall seen in Fig. 5. A similar argument could also be made for energy deposition by alpha particles, but in this case, because the range of alpha particles is much smaller than that of protons, the energy deposition would have to be along the minor axis. For large active volumes of silicon, this key role in the overall energy deposition played by the light reaction products rather than the recoils was earlier investigated via measurements of energy deposition by 14 MeV neutrons impinging on a large volume silicon SBD [ll]. For the 14 MeV neutrons, calculations showed that it was the energy deposition from alpha particles that was the dominant factor.  1819 -I B ,* * t c , With the WNR beam it appears that it is the protons that are primarily responsible. A further verification of this can also be found by examining Table 1 and Figs. 3 and 4, and comparing the energy deposition curves for the virtually identical configurations of 3V, 0 and 3V, 180 in each figure. The beam-device orientation is the same, in one case the beam strikes the front of the device, in the other case, the back. Looking at the two figures, energy deposition for Edep < 5 MeV is identical, small differences show up for 8 MeV < Edep < 5 MeV, and larger differences for Edep > 8 MeV. Considering the very short range of the recoils of < 10 MeV (< 10 pm), especially in the single large volume, it is clear that the recoils within the active volume should produce the same energy deposition in the device whether it is struck from the front or the back. Therefore, it must be the light reaction products with long ranges, protons and alpha particles, that are responsible for the differences in energy deposition. Finally, Fig. 5 shows that energy deposition is greater at 6V bias than at 3V, and the reason for this is the increased depletion depth a the higher voltage. SILICON LARGE OLUMES SBDS) A limited number of charge collection measurements in SBDs by high energy protons have been reported, most by McNulty and co-workers [l]. All these were at the Harvard CycIotron and used SBDs with several different thicknesses of collection depth. Boeing made independent measurements with high energy protons (TIUUMF) and the WNR neutron beam at Los Alamos using SBDs with 10 and 300 pm thickness [2]. Similar measurements were carried out in Russia [3,4]. At the Petersburg Nuclear Physics Institute (PNPI) the energy deposited by a narrow 1 GeV beam of protons (fluence of 4E10 p/cm2) in a 20 pm thick SBD was measured. The differential energy deposition reportd in [3] was converted to the integral deposition in the BGR format. In addition, a larger and thicker SBD, 86 pm, was used by MEPhI personnel and exposed to the 3.65 GeV proton beam (fluence of 3.8E5 p/cm2) at the JINR beam in Dubna, Russia and the energy deposition was measured [4]. This energy deposition was also converted to the BGR format. In Figure 6  we plot the charge deposition curves measured by these four groups: the McNulty group [l], Boeing [2], PNPI [3] and MEPhI [4] for SBDs of various thicknesses. The energy deposition curves appear to fall into three families that mainly depend on the thickness of the SBD: very thin (10 pm), thin (20-25 pm) and thick (90-300 pm). The PNFI 20 pm SBD in a 1 GeV proton beam has a similar energy deposition curve as El-Teleaty's 25 pm detector with 154 MeV protons. Similarly, energy deposition in the MEPhI 90 ~ etector by 3.65 GeV protons matches that of the WNR neutron beam (upper energy of -700 MeV) in the 300 ~ etector. It thus appears that the thickness of the SBD is a more important factor than the Table 2 Characteristics of Various ProtodNeutron SBD Energy Deposition Measurements Shown in Figure 6  I Particle I Part'le I SBD I Symb I Org.or I Ref I I I Energy I Thick I in I Author . I roton I 3600 I90 I IMEPhI 14 I 1E-13 1E-14 1E-15 1E-16 1E-17 1E-18 I Figure 6 Comparison of Energy Deposition in Silicon SBDs from High Energy Protons and Neutrons
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks