New neutron facility for fast neutron irradiation tests of electronics
Collaborative research among Italian and British scientists and engineers from Nast Centre for Nanoscience Nanotechnology and Innovative Instrumentation, CNR (I), ISIS Spallation neutron Source (UK), Universities of Central Lancashire, Padova and Milano Bicocca, and from the avionics and aerospace industries have been using VESUVIO neutron flux…
… and energy spectrum (Figure 1) in benchmark activation measurements to demonstrate that it provides a neutron spectrum similar to the ambient one at sea level, but with an enhancement in intensity of a factor 107 (Figure 2). The VESUVIO beam line at the ISIS spallation neutron source UK Facility over the last few months has been set up for neutron irradiation tests in the neutron energy in the MeV range (press release and ref [16]). The neutron production at ISIS relies upon spallation reactions induced by 800 MeV proton bunches, accelerated through a synchrotron.
Figure 1. Schematic layout of the ISIS facility with the VESUVIO inverse geometry neutron spectrometer
The beam makes about 104 orbits inside the synchrotron as it is accelerated before being kicked in a single revolution into the extracted proton beam line, delivering 4 μC of protons in two 100 ns long pulses to a tungsten spallation target. The entire acceleration process is repeated 50 times per second, so that a mean current of 200 μA is delivered to the target, with a neutron yield of about 30 n/proton. Such conditions are suitable for accelerated testing of electronic components. This was demonstrated by measurements of soft error rates in recent technology FPGAs.
Figure 2: Neutron energy spectrum at VESUVIO (blue line). Also shown for comparison are the neutron energy spectra at LANSCE (red line) and TRIUMF (pink line) facilities, as well as the terrestrial one at sea level (dotted), multiplied by 107 and 108.
The growing availability of integrated circuits featuring minimum dimensions of the order of tens of nanometers, is affecting properties and design methodologies of digital systems. These digital devices are more susceptible to random faults, known as Single Event Effects (SEE), which can occur when a highly energetic particle (such as a neutron present in the environment) causes a disruption of its correct operation by striking sensitive regions of an electronic device [1] (Figure 3).
Figure 3. Example of Industrial sectors potentially affected by SEE - Aerospace: Civilian and military aircraft; Medical:, Implanted electronic devices (pacemakers); Nuclear Industry:, Instrumentation and Control in proximity to reactors; Automotive: Electronics in cars and trains, Signalling and traffic control networks; IT Networks and Telecommunications
SEE have already been identified as a predominant threat to aircraft safety [2] and the effects on electronic components from cosmic radiation is of significant importance for the semiconductors industry [3].The collision process of a neutron with a silicon nucleus, can be either elastic or inelastic. Elastic collisions leave the silicon nucleus intact but they cause it to recoil, thus leaving an intense local ionization trail. At high-energies (in the MeV region) inelastic collisions may lead to a series of direct reactions, called intra-nuclear cascade, which are characterised by the ejection of individual nucleons (protons or neutrons as well as alpha particles) as well as the presence of residual heavy ions that more likely induce SEE into the device (Figure 4).
Figure 4. In 1936 Pfotzer discovered that maximum ionization occurs at about 15 km altitude. Thick atmosphere sustains stable life at sea level. Neutron flux at sea (ground) level: is about 105 neutrons/cm2-year with E>20 MeV
At lower energies, capture reactions may take place with the formation of a compound nucleus which in turn “boils off” nucleons to reach stability. An SEE can be induced when the energy released into the active medium by the charged particles causes the total number of ionization electron-hole pairs collected in a sensitive region of the device to exceed a critical value (which is a characteristic of the device). These radiation-induced errors may be a concern in SRAM-based FPGA’s (Field-Programmable Gate Arrays), where bit-flips in the configuration memory may alter the functionality of the implemented circuits. As a matter of fact, commercial-off-the-shelf devices are becoming popular in mission- or safety-critical applications, since they satisfy the designers’ need for high-performance computing at moderate prices. However, to exploit such devices, fault-tolerant design techniques must be employed, and extensive analyses are needed in order to qualify the robustness of the devices and systems. Experiments with atmospheric neutrons at different altitudes can be carried out but, due to the low intensity, they require very long periods of data acquisition [4]. In this context, neutron sources represent an opportunity due to the availability of high intensity fluxes which allow accelerated irradiation experiments.Presently semiconductor industries perform irradiation tests for example at LANSCE [5] and TRIUMF [6] neutron sources. The Los Alamos Neutron Science Centre (LANSCE) is a multidisciplinary facility for science and technology. The heart of the facility is an 800 MeV high-power linear accelerator that can accelerate both protons and negative hydrogen ions with pulsed beam time patterns that suits a wide variety of experiments. The facility is made of three major experimental areas, of which the Weapons Neutron Research (WNR) is mainly focused to applied research and for testing of semiconductor devices. The WNR uses up to 5 mA of H- beam, chopped and bunched before acceleration to give an adjustable pulse-to-pulse separation, typically 1:8 ms. An un-moderated white neutron beam with useful energies from 100 keV up to above 600 MeV is produced by the interaction of the proton beam with a tungsten target. The neutron beam pulse width is approximately 125 ps , allowing for high-resolution neutron time-of-flight measurements. Irradiation tests of semiconductors are performed at the ICE House (Irradiation of Chips and Electronics) where the neutron spectrum is rather similar to that of neutrons produced in the atmosphere by cosmic rays, but with a neutron flux 108 times higher than the natural one at sea level. This large flux allows users from many countries to use the WNR high-energy-neutron source to characterize electronic components at greatly accelerated rates and study various failure modes caused by neutron irradiation.Irradiation tests of semiconductor devices are also performed at the Neutron Irradiation Facility (NIF) of TRIUMF, Vancouver, Canada. NIF is mainly dedicated to testing avionics and ground-based electronic systems. The neutrons are produced by an intense proton beam from a 500 MeV cyclotron, striking an aluminum beam stop immersed in a tank of cooling water. The neutron beam is then incident in the testing station, where the device under test is lowered into the path of the neutron beam. NIF has an energy spectrum well matched to the atmospheric one, although somewhat softer than that at LANSCE, and can simulate the radiation effects of ten years of atmospheric exposure in just a few minutes. The neutrons produced at TRIUMF have energies up to 400 MeV, with the additional feature that thermal neutrons from the water moderator are also present. Up to now, TRIUMF and LANSCE were the only two facilities worldwide offering such a wide range of neutron energies.The VESUVIO experiments open up the availability of a neutron test facility located in Europe providing increased tests opportunities at different geographical locations for the benefit of those industries for which access to WNR or NIF is not convenient. This work assesses the effectiveness of the ISIS spallation source [7] located at the Rutherford Appleton Laboratory (Chilton, UK), to serve as a unique European facility for accelerated irradiation tests of electronics. This is done by presenting measurements of the neutron spectrum available on the VESUVIO beam line [8] used for this experiment and comparing it to the one of the LANSCE centre. Experimental results obtained in a series of benchmark irradiation experiments on recent technology semiconductors (SRAM-based FPGA) are presented and compared to those obtained at LANSCE.
The VESUVIO beam line has been used for irradiation experiments of electronic chips, taking advantage of the availability of high energy neutrons (in the MeV range). VESUVIO has a primary flight path of 11.055 m (see Figure 1) and a water moderator at ambient temperature that provides a pulsed neutron beam whose spectrum is peaked at about 30 meV and decreases as 1/Eα , with α ≈ 0.9, in the epithermal energy region (above 0.5 eV). The undermoderation of the neutrons results in the presence of an intense flux of neutrons above 1 MeV. The beam diameter at the sample position is about 4.5 cm.
Figure 5: Experimental set up for the tests at the VESUVIO irradiation station at ISIS. For the tests the electronic board is inserted in the sample tank.
In order to evaluate the high energy component of the VESUVIO neutron spectrum, a MCNPX spectrum model was made to simulate the differential neutron flux as a function of energy [9]. This model was benchmarked with two independent measurements (Figure 5). First, neutron activation measurements were performed irradiating selected foils of aluminium, carbon (as graphite), gold and bismuth with the VESUVIO beam for up to 24 hours [10]. The induced activity was measured in a calibrated gamma spectrometer, using two collection times for each foil. This method permitted the identification of the activated nuclide since both the half life and the gamma spectrum were measured. The induced activation for each identified nuclide was corrected for the decay, absorption and dead time, and normalized to the proton current. These results were then compared to the estimated activities derived from the MCNPX model, taking into account the mass of the foil and a start time correction that allows for the decay that occurs in the time taken to move the foils from the VESUVIO beam line to the gamma spectrometer. A consistent ratio between the measured and calculated activity for the range of nuclides gave an integrated intensity above 10 MeV of the VESUVIO beam of (5.82±1.05)×104 nּcm-2ּs-1 at 180 µA proton current and indicated that the MCNPX spectrum was a reasonable model of the differential flux.The second benchmark measurement was based on the event rate in a CCD detector [10a] that placed in the VESUVIO beam line. The number of events above a suitable threshold value (collected charge) in the CCD provides a relative measure of the neutron beam intensity. Direct comparison of the observed event rate with that generated when the same detector was put into the beam at LANSCE provides a value for the LANSCE equivalent flux. The LANSCE equivalent flux is defined as the flux of the LANSCE beam that would be required to induce the same rate of events in the same detector [10b]. This method of characterising a beam takes into account both the energy dependence of the neutron spectrum and also the cross-section for the generation of events above the given threshold level. It therefore provides a measure of the effectiveness of the beam in inducing SEE. Conventionally, SEE cross-sections measured at LANSCE are normalised to the fluence in the LANSCE beam above 10 MeV, even though neutrons below 10 MeV can cause SEE in many devices. Following this convention, LANSCE equivalent flux is also “above 10 MeV”. It is not, however, a direct measure of the received flux above 10 MeV, but a measure of the portion above 10 MeV of the LANSCE beam which would have the same effect as the received flux. The LANSCE equivalent flux at the irradiation position at VESUVIO was measured to be 7.4 × 104 nּcm-2ּs-1 for 180 µA proton current. This means that the fast neutron flux on VESUVIO is 0.18 times as effective at inducing SEE events as the LANSCE beam when operating at 1.8 μA proton current onto the target. Taking account of the softer spectrum at VESUVIO, this independent measurement is consistent with that derived from the activation analysis.
A series of benchmark measurements were performed on semiconductor chips of a recent technology. The aim of these tests was to assess the neutron sensitivity of commercial off-the-shelf SRAM-based Field Programmable Gate Arrays (FPGAs) through a measurement of the static cross section [11]. The static cross section is defined as the ratio of SEE in the configuration memory to the total fluence of hitting neutrons, and represents a generic measure of the FPGA neutron sensitivity. An FPGA is a semiconductor device containing logic components and interconnects that can be programmed “in the field” by the designer, without requiring dedicated manufacturing steps to implement the desired functionality [12]. Applications of FPGAs include a growing number of areas from signal processing to aerospace and defense systems. The logic components inside the FPGA can be used to realize complex combinatorial or sequential functions. The FPGA configuration, i.e. the information describing the implemented circuit, is stored in a memory (configuration memory) which can be of different types: antifuse, flash, or SRAM. If the data in the configuration memory become corrupted, the functionality of the implemented circuit can be altered, possibly leading to computation errors or even functional interrupts. SRAM-based FPGAs are the most versatile and inexpensive devices, but, at the same time, the most sensitive to neutrons and ionizing particles. The measurements on a Xilinx Spartan-3 XC3S200 were carried out by performing a series of short irradiations and the static cross sections, σstatic were computed (Figure 6). The same device was irradiated, in a previous independent experiment, both at LANSCE and with atmospheric neutrons at altitudes of about 1600 m, 3800 m and at sea level [4].
Figure 6. Cross section of SRAMs and FPGA configuration memories as a function of the CMOS technology node as deduced from open literature reports [13-16].
The value of σ static obtained at LANSCE and with natural neutrons were σstatic = 3.43×10-14 cm2/bit and σstatic = 3.17×10-14 cm2/bit, respectively. The value obtained at ISIS is σstatic = (3.0±0.6) ×10-14 cm2/bit . This result is in good agreement with the results obtained by the other two independent measurements. As LANSCE is considered the standard site for this kind of tests, the results from VESUVIO at ISIS indicate its usefulness as an irradiation facility. Work is in progress to improve the sample environment for this kind of study. In the longer term, it is planned to design and build a new irradiation beam line with improved spectral conditions.
The co-authors on the study:
C. Andreani, M. P. De Pascale, P. Picozza, A. Pietropaolo, A. Salsano, Centro NAST, Università degli Studi di Roma Tor Vergata, Italy G. Gorini, M. Tardocchi, Dipartimento di Fisica “G. Occhialini”, Università degli Studi di Milano-Bicocca, Italy A. Paccagnella, S. Gerardin, Dipartimento di Ingegneria dell’Informazione, Università di Padova, Italy C. D. Frost, S. Ansell, ISIS facility, Rutherford Appleton Laboratory, Great Britain S. P. Platt, Department of Technology, University of Central Lancashire, Great Britain
References
[1] Ziegler J F, Lanford WA, J. Appl. Phys. 52(6) 4305, (1981); Normand E., IEEE Trans. Nucl. Sci. 43(6), 2742 (1996).
[2] T. J. O’Gorman, J. M. Ross, A. H. Taber, J. F. Ziegler, H. P. Muhlfeld, I C. J. Montrose, H. W. Curtis, J. L. Walsh, IBM I. RES. DEVELOP. 40 NO. 1, JANUARY 1996.
[3] R. C. Baumann, IEEE Trans. Dev. Mat. Rel. 1, 17 (2001).
[4] A. Lesea, S. Drimer, J.J. Fabula, C. Carmichael, P. Alfke, IEEE Trans. Dev. Mat. Rel. 5, 317 (2005).
[6] http://www.triumf.ca/welcome/index.html.
[7] http://www.isis.rl.ac.uk/.[8] http://www.fisica.uniroma2.it/~vesuvio/dins/index.html; R. Senesi, C. Andreani, Z. Bowden, D. Colognesi, E. Degiorgi. A.L. Fielding, J. Mayers, M. Nardone, J. Norris, M. Praitano, N.J. Rhodes, W.G. Stirling, J. Tomkinson, C. Uden, Physica B 276-278, 200 (2000).
[9] http://mcnpx.lanl.gov/.[10] C. Frost, ISIS Internal report (2006)
[10a] S. P. Platt, B. Cassels, Z. Török, J. Phys. Conf. Ser. 15 (2005) 172–176 [10b] S. P. Platt, Z. Török, IEEE Trans. Nuc. Sci. 54 (4) 2007 in press.
[11] M. Violante, L. Sterpone, A. Manuzzato, S. Gerardin, P. Rech, M. Bagatin, A. Paccagnella, C. Andreani, G. Gorini, A. Pietropaolo, G. Cardarilli, S. Pontarelli, C. Frost, paper in publication on IEEE Trans. Nucl. Science (2007).
[12] http://www.xilinx.com/prs_rls/design_win/0412_marsrover.htm; J. Rose, A. El Gamal, A. Sangiovanni-Vincetelli, IEEE procedings, 81, no. 7, 1013, (1993).
[13] T. Granlund et al., IEEE Trans. Nuc. Sci., 50, No. 6, 2065 (2003).
[14] P. Roche et al., IEEE Trans. Nuc. Sci., 50, No. 6, 2046, (2003).
[15] J. Baggio et al., IEEE Trans. Nuc. Sci., 51, No. 6, 3420, (2004).[16] D. Lambert et al., IEEE Trans. Nuc. Sci., 51, No. 6, 3435, (2004).
[16] C. Andreani, A. Pietropaolo, A. Salsano, G. Gorini, M. Tardocchi, A. Paccagnella, S. Gerardin, C. D. Frost, S. Ansell, and S. P. Platt, Facility for fast neutron irradiation tests of electronics at the ISIS spallation neutron source, Applied Physics Letters, 92, 114101 (2008).