Much of the experimental part of this work was made possible by the help and friendship of my colleague Nathaniel Dodds. Of the types of radiation e↵ects plaguing modern-day electronic devices, single-event e↵ects (SEE - radiation e↵ects caused by a single particle impact) have become increasingly important. This dissertation presents new research that furthers the understanding of the mechanisms behind SEUs in modern equipment, both for equipment exposed to the natural space and ground-level radiation environments.
One of the most prominent alternatives is the use of a 14 MeV monoenergetic neutron beam generated by a deuterium and tritium fusion reaction [13]. In [12], a comparison of the SEU cross sections measured using a 14 MeV neutron source and the WNR neutron spectrum was made. For this reason, 14 MeV neutrons under predict the WNR MCU neutron cross section for the device considered here.
BACKGROUND
The radiation environment outside Earth's atmosphere contains a wide range of particle types and energies, as shown in Table 1 [16]. The types of radiation that exist are: 1) trapped particles in the Earth's magnetic field (which form the Van Allen belts), 2) energetic particles emitted by the Sun (solar particles), and 3) particles originating from outside our solar system (galactic radiation). cosmic rays (GCRs)). The heavy ion population in the GCR spectra is represented by most elements in the periodic table, but only elements up to and including iron are found in significant amounts (see Figure 1A) [17].
Energetic heavy ions (Z > 2) and alpha particles are rare compared to protons and account for only about 1% of the total GCR flux. At sea level, the electronics are continuously bombarded with a flux of energetic neutrons spanning a wide range of energies as shown in Fig. 4 shows some of the different classifications of SEEs, hard and soft errors, that are currently researched.
SBU!
SEE!
MBU/MCU!
SET!
SEL!
SEB!
SEGR!
SHE!
The location of the alpha peak will vary depending on the amplifier setting in the PHA system. In this work, all PHA data are analyzed in the form of an integral cross-section. Due to the size of the diodes, only four diodes were mounted in each DIP.
Like experiments using proton radiation, all recorded charge accumulation events were due to neutron-nucleus collisions. This section discusses the experimental details of the TOF setup, energy calculation, and limitations in TOF measurements. The start pulse for TAC is provided by the facility that signals the time of neutron generation.
The amplitude of the pulse from the TAC is then proportional to the time between neutron generation and the charge collection event. This section contains a description of the Monte Carlo simulation tool used and how it was used in this work.
MECHANISMS OF PROTON-INDUCED SINGLE EVENT EFFECTS
In this way, the dependence of the charge collection cross section on the proton energy and angle can be analyzed as a function of the smallest collected charge. This charge collection cross section is analogous to the cross section of an SEU device with the same dimensions of the sensitive volume. However, a qualitative understanding of the proton-induced SEU cross section can be obtained from these data.
23, the charge collection cross section increases with proton energy for the diodes with SiO2 overlays for any Qmin. 23 also shows that, for 200 and 500 MeV protons, the presence of W near the sensitive volume region of a device causes a significant increase in the cross section for charge collection events above ⇠0.5 pC. Monte Carlo simulations confirm that proton-induced fission fragments of W, and not simply proton-induced spallation fragments, are the cause of the increased charge collection cross section shown in Fig.
23 shows that the dependence of the charge collection cross section on proton energy varies with Qmin. For low charge collection events of 0.2 pC, the charge collection cross section measured with 500 MeV protons is only about 30% larger than with 200 MeV protons. Due to the thin charge collection region in these diodes, a significant increase in the charge collection cross section at grazing angles can be expected.
A significant increase in the charge collection cross section of the diodes would probably be observed if they were irradiated at an angle at lower proton energies. Note that although this model includes proton-induced W decay, it does not show any predicted difference in the collected charge cross section for a device with W. This means that high-Z materials can increase the proton-induced SEU cross section for high-energy protons and high-critical devices charge.
Monte Carlo simulations show that high-LET secondary particles from a proton-W collision originate from proton-induced fission.
![Figure 22 Figure from Howe et al. [5] showing no increase in proton-induced charge collection for devices with a W and an oxide overlayer.](https://thumb-ap.123doks.com/thumbv2/123dok/10730964.0/60.918.300.680.155.451/figure-figure-showing-increase-induced-collection-devices-overlayer.webp)
MECHANISMS OF NEUTRON-INDUCED SINGLE EVENT EFFECTS
31 shows that for a minimum collected charge, Qmin, less than 0.6 pC, the e↵ect of the W layer on the charge collection cross section is smaller. However, for high Qmin, the presence of W can significantly increase the charge collection cross section. To reduce the running time of the simulations, cross-sectional biasing techniques were used in MRED (see [43,68]).
Comparisons of the cross-section curves of these two overlay configurations are shown as a function of Qcrit in Fig. Figure 35 shows that the W overlayer will affect the MCU cross-section more significantly for higher MCU multiplicity. 35 shows that the W overlayer has no significant effect on the MCU cross section for MCU events affecting 2-4 bits at this Qcrit.
36 shows that for Qcrit <27 fC the SBU cross-section curves for the 14 MeV neutrons and the WNR neutron spectrum agree within a factor of 2. 36 also shows that the 2BU cross-section curves agree within a factor of 2 only for Qcrit < 1.2 fC and that 14 MeV neutrons significantly underestimate the 2BU WNR response for higher Qcrit values. 37 shows that for 3+BU events the 14 MeV neutrons drastically underestimate the MCU cross section except at extremely low (< 0.6 fC) Qcrit values.
The 28Si(n, n +↵)24Mg reaction also contributes to the MCU events, however the cross section for this reaction is approximately one order. However, it is important to note that for an SRAM with Qcrit of about 1 fC, the MCU cross section for 14 MeV neutrons and the WNR agree by a factor of 2. To investigate the role that neutron energy plays in the MCU response to the neutron spectrum WNR, simulation data from the 32⇥32-bit SRAM structure with only silicon overlays was presented in a 3-D cross-section plot in Fig.
43 shows an increase in cross-section with high MCU number for the W-overlayer SRAM, consistent with Fig.
CONCLUSIONS
As semiconductor devices become smaller and smaller, these results will provide insight that will aid in device testing and understanding future observations of radiation effects.
APPENDIX A
LINEAR ENERGY TRANSFER
This appendix briefly covers the basic physics involved in the energy loss calculations and the limitations of the approximations and assumptions made. For energetic ions, the energy loss in materials generally falls along a straight path, and most of the lost energy is transferred to the electrons in the material. Here p, E and v are the initial momentum, energy and velocity of the incident ion and the primed variables are after the collision.
The equation is best put in terms of momentum transfer q= p p0, since this is related to kinetic energy. For highQ, the electron binding energy is neglected and the initial and final electron states are taken as free electrons. The I factor is the average excitation energy of atomic electrons specific to a material.
While this assumption holds for most electrons, it fails for some of the more tightly bound electrons. The region 1 assumption leading to exponential expansion also fails ifq·r is not much smaller than ¯h. Since the other factor in the Born expansion contains a term proportional to the cube of the incident particle charge, charged pions and heavy ions deviate from the predictions of Eq.
L2 ( ) has changed over the years to a more empirical form [89], but remains part of the general LET equation. This is a more serious consideration for thin volumes, since fluctuations in the stored energy are a larger percentage of the total stored energy. Experimental results for LET in bulk materials can vary between experiments by several percent due to differences in the actual atomic structure of the materials.
The validation and adjustment of the correction factors in the LET theory is an ongoing work, however, the overall accuracy of the theory was very good with an average agreement with experimental data of less than 5% [65].
APPENDIX B
MRED STANDARD MODE EXAMPLE CODE
APPENDIX C
MRED SINGLE EVENT MODE EXAMPLE CODE
BIBLIOGRAPHY
E↵ffects of neutron energy and high-z materials on single-event reversals and multiple cell disorders," IEEE Trans. Dyer, "Radiation ↵efects in the new millennium - old realities and new issues" , IEEE NSREC Short Course, 1998. King, “Charge Generation from Secondary Particles by Nuclear Reactions in Beol Materials,” IEEE Trans.
Lacoe, “Radiation Effects in Advanced Commercial Technologies: How Device Scaling Has Influenced Space Electronics Choices,” IEEE NSREC Short Course, 2003. Rivet, “Consequences of Spatial Dependence of Disturbances in Single-Event Threshold in Shame Measured with a Pulsed Laser,” IEEE Trans. . Barnaby, “Analysis of the effect of mos device geometry on predicted seu cross sections,” IEEE Trans.
Seidleck, “Effect of Low-Energy Proton-Induced Interference on Test Methods and Rate Predictions,” IEEE Trans. Massengill, “Implications of Nuclear Reactions for Test Methods and Single Event Effect Analysis,” IEEE Trans. Fabula, “Monte-carlo based orbital excitation rate prediction for latch-hardened radiation,” IEEE Trans.
Validation of Nuclear Reaction Codes for Proton-Induced Radiation Effects: The Case for cem03," IEEE Trans. Mendenhall, "Enhancement of the Multibit Inversion Rate by Neutrons at Large Incidence Angles," IEEE Trans. Palau, "Simulations Monte-carlo to quantify neutron-induced multiple bit perturbations in advanced srams,” IEEE Trans.
Normand, “Extensions of the burst generation rate method for wider application to single proton/neutron induced event e↵ects,” IEEE Trans.
![Figure 2 Measured terrestrial neutron energy spectrum [7] plotted with the published JEDEC fit to the data [24].](https://thumb-ap.123doks.com/thumbv2/123dok/10730964.0/18.918.228.721.152.467/figure-measured-terrestrial-neutron-energy-spectrum-plotted-published.webp)
![Figure 11 Cosmic-ray trajectories recorded by nuclear photographic emul- emul-sions flown in space and subsequently developed: A) track of heavily ionizing particle, B) inelastic nuclear reaction and secondary particle tracks [53].](https://thumb-ap.123doks.com/thumbv2/123dok/10730964.0/31.918.228.743.151.353/trajectories-recorded-photographic-subsequently-developed-ionizing-inelastic-secondary.webp)
