- Email: [email protected]
Single event effects rate predictions in space
1260
Nuclear
Instruments
and Methods
in Physics Research
B56/57
(1991) 1260-1262 North-Holland
Single event effects rate predictions
in space
John R. Letaw Sevem Communications
Corporation, 223 Benfield Park Drive, Millersville,
MD 21108, USA
Computational methods for estimating single event error rates in space are reviewed. Single event effects are a source of error in spacecraft microelectronics caused by the passage of high-energy charged particles such as cosmic rays. These one- or few-bit errors may cause data loss or system malfunctions without any permanent damage to the device. Designers must be assured that errors fall below tolerable limits for their mission. Susceptibility of microelectronic devices is characterized by exposure to proton and heavy ion beams in laboratory facilities. The behavior of devices in space is generally estimated using computational models in conjunctio: with the accelerator data. Models of the space radiation environment include the CREME (cosmic ray effects on microelectronics) model of the galactic cosmic ray environment, the AP-8 trapped proton model, and various models of solar flare events. Radiation transport codes determine changes in the radiation environment after passage through the Barth’s magnetic field and spacecraft structural materials. Semiconductor device models are used to convert particles fluxes into error rates.
1. Introduction Single event effects when
individual
“cosmic
in spacecraft ray”
electronics
particles
pass
occur
through
a
The charged particle removes electrons from the atoms along its path creating an increased density of electron-hole pairs. The resulting free charge causes a current pulse in the circuit which can switch bistable elements, changing a 1 into a 0, or vice versa. Single event effects differ from other radiation effects because they generally do not cause significant permanent damage to the device. Furthermore, they stem from the passage of a single particle, not the combined effect of thousands of particles over an extended period. Single event effects obviously can have great impact on space missions. If the telemetry system is affected, data can be lost. If detector systems are affected, false readings can occur. If control systems are affected, a spacecraft can become unmanageable. It has been recognized for a decade or more that spacecraft designers must be cognizant of single event effects. Engineers must address two separate issues to determine the hardness of electronics systems against single event effects. First, individual devices must be tested to characterize their sensitivity to radiation which occurs in space. Second, these data are used along with models of the space radiation environment to estimate the behavior of the system in its intended orbit. The objective is to predict a single-event upset (SEU) or error rate in the device. In this paper we address the second of these tasks, that is, numerical modeling of the effects of the space radiation environment on semiconductors. The many sensitive
semiconductor
0168-583X/91/$03.50
memory
element.
0 1991 - Elsevier Science Publishers
aspects of this modeling include identifying and modeling the sources of radiation, specifying the orbital trajectory of the spacecraft, specifying the distribution of spacecraft structural materials, determining the attenuation of the radiation in the Earth’s magnetic field and the spacecraft structure, and using device models to estimate the effects of radiation. Each of these aspects of single event effects modeling will be addressed in this paper. In the summary existing codes for performing single event effects modeling will be mentioned, along with the outlook for future work.
2. The natural radiation environment There are three major sources of ionizing radiation in space. Each of these sources contains electrons, protons, and heavy ions with energies in the range from 1 MeV to several GeV and higher. Heavy ions, for example iron nuclei, are the principal cause of single event effects. Protons and electrons are a direct cause of upsets only in very sensitive devices. Indirectly protons may cause an upset via nuclear interactions in which substrate nuclei recoil with several MeV of kinetic energy. Per proton, secondary processes are much less likely to cause an upset than a single heavy ion; however, there are far more protons than heavy nuclei in space environments. High-energy ions are prominent in the galactic cosmic radiation (GCR). This radiation comes from the stars with energies averaging a few GeV per nucleon. GCR is attenuated by solar and terrestrial magnetic fields, a factor discussed in section 3. GCR is a relatively lowlevel source of heavy ions which varies only slowly over
B.V. (North-Holland)
1261
J.R. Letaw / Single event effects rate predictions in space the eleven-year solar cycle. Because of their energy, shielding against GCR is usually impractical. The CREME (cosmic ray effects on microelectronics) model of GCR [l] has been widely accepted in the aerospace community. Three GCR environments are modeled: solar maximum, solar minimum, and 90% worst-case. Solar activity slows the GCR flux so that the solar minimum environment is more intense than the solar maximum environment. The 90% worst-case environment provides an upper limit of GCR intensity exceeded only 10% of the time. This model is often used by engineers as a baseline for single event effects computations. GCR coming from outside the magnetosphere are attenuated as they pass through Earth’s magnetic field. Over equatorial regions, particles under a few GeV are deflected away from spacecraft orbits. These orbits are protected from all but the highest energy GCR. Over polar regions, particles move along field lines with little deflection and spacecraft orbits are open to the full radiation fluxes from outside. These processes are modeled in a second CREME report [2]. Trapped protons in the Van Allen belts are a second source of radiation in space. These protons are confined to specific areas of the Earth’s magnetic field and hence are not a concern in many space systems (particularly geosynchronous satellites). In low-Earth orbit, the radiation belts dip near Earth in th South Atlantic anomaly (SAA) region near Brazil. The intensity is much higher in the SAA than elsewhere at the same altitude. Trapped protons may have energies up to a few hundred MeV. High-energy particle intensity in the Van Allen belts is greatest at about 2000 km altitude. The authoritative model of the trapped protons is the AP-8 model [3]. Trapped proton environments at solar maximum and solar minimum are represented. The model is parameterized in terms of B and L coordinates which are adapted to the Earth’s magnetic field. Using the variety of magnetic field models contained in ALLMAG [4], one may deduce the trapped proton flux at any geographical location in space. In ref. [3] the authors point out that the AP-8 model requires improvement to represent low-altitude fluxes, as well as fluctuations during magnetically active periods. The third source of ionizing radiation in space is solar energetic particles (SEP). During much of the year there are active areas on the sun emitting radiation of all types. Occasionally, and at irregular intervals, energetic ions are emitted which may affect spacecraft electronics. Every decade or so there is an extremely large flare with intense particle levels which dominate the space radiation environment. The average particle energy in these flares is generally 100 MeV or less, but some proton energies may exceed 1 GeV. Ring [5] has performed an extensive statistical analysis of solar flare frequency and proton intensity. These
models have been incorporated into the SOLPRO program [6]. The CREME model [l] contains models of the peak proton and heavy ion intensity during several classes of solar energetic particle events. These models are very useful for establishing worst-case scenarios for spacecraft electronics. Solar flare intensities are drastically reduced by the Earth’s magnetic shielding. In low-Earth orbits of 28.5 o (a typical shuttle inclination) very few, if any, energetic particles from the sun penetrate the field. In high-inclination orbits (including polar orbits) particles may penetrate the field freely above about 60° latitude.
3. Radiation transport Once the desired space radiation environment is chosen, it is necessary to estimate how the radiation flutes will change in passing from outside the spacecraft to inside the spacecraft. The principal mechanism for attenuation of radiation in spacecraft materials is ionization loss. Ionizing radiation deposits energy in materials through collisions with bound electrons. As the primary particles lose energy to the material they pass through, they slow down. The slowing down is continuous for high-energy particles and is represented by a stopping power or LET (linear energy transfer) function which depends on the energy and charge of the ion. If the differential particle flux is F( E, x) at energy E after passing through material thickness X, and the stopping power is w(E) then S’(E,x)/ax=a(w(E)
F(E,x))/Z.
(1)
The CREME model contains computer codes for solving eq. (1) for all heavy ion species. Another mechanism of radiation attenuation is nuclear spallation. These reactions generally break up heavy ions into lighter pieces with lower charge. The lower charged nuclei are less effective at causing single event effects. The effect of nuclear spallation reactions on GCR fluxes is modeled by the UPROP code [7]. Additional particles, such as protons, neutrons, alphas, and pions, are often emitted from a nuclear interaction site. The potential of these secondary particles for causing single event effects is a subject of current research.
4. Device models After the radiation environment inside a spacecraft is established, one may use that information to estimate SEU rates in semiconductor devices. Through accelerator measurements the upset cross section u (typically having units of cm’) is determined for a variety of radiation types. For SEUs caused by direct ionization, the LET or stopping power of the incident heavy ion XVIII. SOFT UPSETS
1262
J. R. L-etaw / Single event effects race predictions in space
characterizes the radiation. For proton-induced SEUs, the proton energy characterizes the radiation. Pickel and Blandford [8] have proposed a simple model of a semiconductor memory element for SEU estimates. The memory element is assumed to be a rectangular parallelepiped. For a normally incident accelerator beam, the surface area of the parallelepiped is equal to the upset cross section divided by the number of bits in the device. The LET threshold (Lr) is the minimum particle stopping power which can cause an error (usually in units of MeV/(mg/cm’). Using an estimate of the device thickness r (usually in units of t.tm) one may determine that the critical charge (Q,-) required to cause a upset is Q&C)
= 2.27.1+~,
(2)
where p is the density of silicon and a factor of 22.7 converts from MeV of energy deposited to pC of charge deposited in silicon. For a particle passing through the parallelepiped at any point one may compute the chord length, use the LET to compute the energy deposited, and use the factor of 22.7 to compute the charge deposited. If the charge deposited exceeds Q, an error will occur. Bendel and Petersen [9] have deviced a model for proton-induced upsets in silicon. The mechanism for these upsets requires a nuclear interaction between the incident proton and a silicon atom in the substrate. The silicon atom recoiis with an energy of a few MeV. The recoil nucleus deposits charge locally by ionization. An upset cross section is measured at a single proton energy. This measurement is associated with a set of empirically determined curves which contain the energy dependence of the upset cross section. Armed with the upset cross section as a function of energy and the incident proton spectrum, the SEU rate may be computed directly. There is no reference to bit geometry in this model because each recoil nucleus has a short range and deposits much of its energy in the region of a single bit.
5. Software tools and future work The software tools used to estimate single-event upset rates are soid by two sources. The original CREME model is available from the NOAA National Geophysi-
cal Data Center. This FORTRAN code contains the CREME GCR and SEP models, a radiation transport code for alumni shields, and two upset models for silicon devices. Recently Sevem Communications Corporation has introduced a product called SPACE RADIATION which is based on CREME. In addition to the capabilities of CREME, SPACE RADIATION contains the trapped proton modeling tools; procedures for transport through water, plastic, and other materials; and upset modeling for GaAs devices. SPACE RADIATION performs input data checking, has on-line help, and includes a database management system and report generation for documenting computations. There are many areas in which the basic SEU modeling tools can be improved. Continued efforts to compare the space radiation environment models with observations will provide us with the assurance that SEU predictions are reasonable for spacecraft engineering purposes. Extensions of the modeling capability to other ionizing radiation types and to aircraft systems will make the tools comprehensive. Further efforts at detailed modeling of semiconductor devices will improve our understanding of SEU mechanisms and ultimately lead to better device models.
References
111J.H. Adams, Jr., R. Silberberg and C.H. Tsao, NRL Memorandum Report 4506 (Naval Research Laboratory, Washington, DC, 1981). [21 J.H. Adams, Jr., J.R. Letaw and D.F. Smart, NRL Memorandum Report 5402 (Naval Research Laboratory, Washington, DC, 1983). 131 D.M. Sawyer and J.I. Vette, NASA-TM-X-72605 (National Aeronautics and Space Administration, Washington, DC, 1976). 141 E.G. itassinopoulos and G.D. Mead, NASA-TM-X-65844 (National Aeronautics and Space Administration, Washington, DC, 1972). PI J.H. King, J. Spacecraft 11 (1974) 401. 161 E.G. Stassinopoulos, NSSDC 75-11 (National Space Science Data Center, Greenbelt, MD, 1975). [71 J.R. Letaw, SCC Report 89-02 (Sevem Communications Corporation, Millersville, MD, 1989). PI J.C. Pickel and J.T. Blandford, Jr., IEEE Trans. Nucl. Sci. NS-25 (1978) 1166. A W.L. Bendel and E.L. Petersen, IEEE Trans. Nud. Sci. NS-30 (1983) 4481.
Nuclear
Instruments
and Methods
in Physics Research
B56/57
(1991) 1260-1262 North-Holland
Single event effects rate predictions
in space
John R. Letaw Sevem Communications
Corporation, 223 Benfield Park Drive, Millersville,
MD 21108, USA
Computational methods for estimating single event error rates in space are reviewed. Single event effects are a source of error in spacecraft microelectronics caused by the passage of high-energy charged particles such as cosmic rays. These one- or few-bit errors may cause data loss or system malfunctions without any permanent damage to the device. Designers must be assured that errors fall below tolerable limits for their mission. Susceptibility of microelectronic devices is characterized by exposure to proton and heavy ion beams in laboratory facilities. The behavior of devices in space is generally estimated using computational models in conjunctio: with the accelerator data. Models of the space radiation environment include the CREME (cosmic ray effects on microelectronics) model of the galactic cosmic ray environment, the AP-8 trapped proton model, and various models of solar flare events. Radiation transport codes determine changes in the radiation environment after passage through the Barth’s magnetic field and spacecraft structural materials. Semiconductor device models are used to convert particles fluxes into error rates.
1. Introduction Single event effects when
individual
“cosmic
in spacecraft ray”
electronics
particles
pass
occur
through
a
The charged particle removes electrons from the atoms along its path creating an increased density of electron-hole pairs. The resulting free charge causes a current pulse in the circuit which can switch bistable elements, changing a 1 into a 0, or vice versa. Single event effects differ from other radiation effects because they generally do not cause significant permanent damage to the device. Furthermore, they stem from the passage of a single particle, not the combined effect of thousands of particles over an extended period. Single event effects obviously can have great impact on space missions. If the telemetry system is affected, data can be lost. If detector systems are affected, false readings can occur. If control systems are affected, a spacecraft can become unmanageable. It has been recognized for a decade or more that spacecraft designers must be cognizant of single event effects. Engineers must address two separate issues to determine the hardness of electronics systems against single event effects. First, individual devices must be tested to characterize their sensitivity to radiation which occurs in space. Second, these data are used along with models of the space radiation environment to estimate the behavior of the system in its intended orbit. The objective is to predict a single-event upset (SEU) or error rate in the device. In this paper we address the second of these tasks, that is, numerical modeling of the effects of the space radiation environment on semiconductors. The many sensitive
semiconductor
0168-583X/91/$03.50
memory
element.
0 1991 - Elsevier Science Publishers
aspects of this modeling include identifying and modeling the sources of radiation, specifying the orbital trajectory of the spacecraft, specifying the distribution of spacecraft structural materials, determining the attenuation of the radiation in the Earth’s magnetic field and the spacecraft structure, and using device models to estimate the effects of radiation. Each of these aspects of single event effects modeling will be addressed in this paper. In the summary existing codes for performing single event effects modeling will be mentioned, along with the outlook for future work.
2. The natural radiation environment There are three major sources of ionizing radiation in space. Each of these sources contains electrons, protons, and heavy ions with energies in the range from 1 MeV to several GeV and higher. Heavy ions, for example iron nuclei, are the principal cause of single event effects. Protons and electrons are a direct cause of upsets only in very sensitive devices. Indirectly protons may cause an upset via nuclear interactions in which substrate nuclei recoil with several MeV of kinetic energy. Per proton, secondary processes are much less likely to cause an upset than a single heavy ion; however, there are far more protons than heavy nuclei in space environments. High-energy ions are prominent in the galactic cosmic radiation (GCR). This radiation comes from the stars with energies averaging a few GeV per nucleon. GCR is attenuated by solar and terrestrial magnetic fields, a factor discussed in section 3. GCR is a relatively lowlevel source of heavy ions which varies only slowly over
B.V. (North-Holland)
1261
J.R. Letaw / Single event effects rate predictions in space the eleven-year solar cycle. Because of their energy, shielding against GCR is usually impractical. The CREME (cosmic ray effects on microelectronics) model of GCR [l] has been widely accepted in the aerospace community. Three GCR environments are modeled: solar maximum, solar minimum, and 90% worst-case. Solar activity slows the GCR flux so that the solar minimum environment is more intense than the solar maximum environment. The 90% worst-case environment provides an upper limit of GCR intensity exceeded only 10% of the time. This model is often used by engineers as a baseline for single event effects computations. GCR coming from outside the magnetosphere are attenuated as they pass through Earth’s magnetic field. Over equatorial regions, particles under a few GeV are deflected away from spacecraft orbits. These orbits are protected from all but the highest energy GCR. Over polar regions, particles move along field lines with little deflection and spacecraft orbits are open to the full radiation fluxes from outside. These processes are modeled in a second CREME report [2]. Trapped protons in the Van Allen belts are a second source of radiation in space. These protons are confined to specific areas of the Earth’s magnetic field and hence are not a concern in many space systems (particularly geosynchronous satellites). In low-Earth orbit, the radiation belts dip near Earth in th South Atlantic anomaly (SAA) region near Brazil. The intensity is much higher in the SAA than elsewhere at the same altitude. Trapped protons may have energies up to a few hundred MeV. High-energy particle intensity in the Van Allen belts is greatest at about 2000 km altitude. The authoritative model of the trapped protons is the AP-8 model [3]. Trapped proton environments at solar maximum and solar minimum are represented. The model is parameterized in terms of B and L coordinates which are adapted to the Earth’s magnetic field. Using the variety of magnetic field models contained in ALLMAG [4], one may deduce the trapped proton flux at any geographical location in space. In ref. [3] the authors point out that the AP-8 model requires improvement to represent low-altitude fluxes, as well as fluctuations during magnetically active periods. The third source of ionizing radiation in space is solar energetic particles (SEP). During much of the year there are active areas on the sun emitting radiation of all types. Occasionally, and at irregular intervals, energetic ions are emitted which may affect spacecraft electronics. Every decade or so there is an extremely large flare with intense particle levels which dominate the space radiation environment. The average particle energy in these flares is generally 100 MeV or less, but some proton energies may exceed 1 GeV. Ring [5] has performed an extensive statistical analysis of solar flare frequency and proton intensity. These
models have been incorporated into the SOLPRO program [6]. The CREME model [l] contains models of the peak proton and heavy ion intensity during several classes of solar energetic particle events. These models are very useful for establishing worst-case scenarios for spacecraft electronics. Solar flare intensities are drastically reduced by the Earth’s magnetic shielding. In low-Earth orbits of 28.5 o (a typical shuttle inclination) very few, if any, energetic particles from the sun penetrate the field. In high-inclination orbits (including polar orbits) particles may penetrate the field freely above about 60° latitude.
3. Radiation transport Once the desired space radiation environment is chosen, it is necessary to estimate how the radiation flutes will change in passing from outside the spacecraft to inside the spacecraft. The principal mechanism for attenuation of radiation in spacecraft materials is ionization loss. Ionizing radiation deposits energy in materials through collisions with bound electrons. As the primary particles lose energy to the material they pass through, they slow down. The slowing down is continuous for high-energy particles and is represented by a stopping power or LET (linear energy transfer) function which depends on the energy and charge of the ion. If the differential particle flux is F( E, x) at energy E after passing through material thickness X, and the stopping power is w(E) then S’(E,x)/ax=a(w(E)
F(E,x))/Z.
(1)
The CREME model contains computer codes for solving eq. (1) for all heavy ion species. Another mechanism of radiation attenuation is nuclear spallation. These reactions generally break up heavy ions into lighter pieces with lower charge. The lower charged nuclei are less effective at causing single event effects. The effect of nuclear spallation reactions on GCR fluxes is modeled by the UPROP code [7]. Additional particles, such as protons, neutrons, alphas, and pions, are often emitted from a nuclear interaction site. The potential of these secondary particles for causing single event effects is a subject of current research.
4. Device models After the radiation environment inside a spacecraft is established, one may use that information to estimate SEU rates in semiconductor devices. Through accelerator measurements the upset cross section u (typically having units of cm’) is determined for a variety of radiation types. For SEUs caused by direct ionization, the LET or stopping power of the incident heavy ion XVIII. SOFT UPSETS
1262
J. R. L-etaw / Single event effects race predictions in space
characterizes the radiation. For proton-induced SEUs, the proton energy characterizes the radiation. Pickel and Blandford [8] have proposed a simple model of a semiconductor memory element for SEU estimates. The memory element is assumed to be a rectangular parallelepiped. For a normally incident accelerator beam, the surface area of the parallelepiped is equal to the upset cross section divided by the number of bits in the device. The LET threshold (Lr) is the minimum particle stopping power which can cause an error (usually in units of MeV/(mg/cm’). Using an estimate of the device thickness r (usually in units of t.tm) one may determine that the critical charge (Q,-) required to cause a upset is Q&C)
= 2.27.1+~,
(2)
where p is the density of silicon and a factor of 22.7 converts from MeV of energy deposited to pC of charge deposited in silicon. For a particle passing through the parallelepiped at any point one may compute the chord length, use the LET to compute the energy deposited, and use the factor of 22.7 to compute the charge deposited. If the charge deposited exceeds Q, an error will occur. Bendel and Petersen [9] have deviced a model for proton-induced upsets in silicon. The mechanism for these upsets requires a nuclear interaction between the incident proton and a silicon atom in the substrate. The silicon atom recoiis with an energy of a few MeV. The recoil nucleus deposits charge locally by ionization. An upset cross section is measured at a single proton energy. This measurement is associated with a set of empirically determined curves which contain the energy dependence of the upset cross section. Armed with the upset cross section as a function of energy and the incident proton spectrum, the SEU rate may be computed directly. There is no reference to bit geometry in this model because each recoil nucleus has a short range and deposits much of its energy in the region of a single bit.
5. Software tools and future work The software tools used to estimate single-event upset rates are soid by two sources. The original CREME model is available from the NOAA National Geophysi-
cal Data Center. This FORTRAN code contains the CREME GCR and SEP models, a radiation transport code for alumni shields, and two upset models for silicon devices. Recently Sevem Communications Corporation has introduced a product called SPACE RADIATION which is based on CREME. In addition to the capabilities of CREME, SPACE RADIATION contains the trapped proton modeling tools; procedures for transport through water, plastic, and other materials; and upset modeling for GaAs devices. SPACE RADIATION performs input data checking, has on-line help, and includes a database management system and report generation for documenting computations. There are many areas in which the basic SEU modeling tools can be improved. Continued efforts to compare the space radiation environment models with observations will provide us with the assurance that SEU predictions are reasonable for spacecraft engineering purposes. Extensions of the modeling capability to other ionizing radiation types and to aircraft systems will make the tools comprehensive. Further efforts at detailed modeling of semiconductor devices will improve our understanding of SEU mechanisms and ultimately lead to better device models.
References
111J.H. Adams, Jr., R. Silberberg and C.H. Tsao, NRL Memorandum Report 4506 (Naval Research Laboratory, Washington, DC, 1981). [21 J.H. Adams, Jr., J.R. Letaw and D.F. Smart, NRL Memorandum Report 5402 (Naval Research Laboratory, Washington, DC, 1983). 131 D.M. Sawyer and J.I. Vette, NASA-TM-X-72605 (National Aeronautics and Space Administration, Washington, DC, 1976). 141 E.G. itassinopoulos and G.D. Mead, NASA-TM-X-65844 (National Aeronautics and Space Administration, Washington, DC, 1972). PI J.H. King, J. Spacecraft 11 (1974) 401. 161 E.G. Stassinopoulos, NSSDC 75-11 (National Space Science Data Center, Greenbelt, MD, 1975). [71 J.R. Letaw, SCC Report 89-02 (Sevem Communications Corporation, Millersville, MD, 1989). PI J.C. Pickel and J.T. Blandford, Jr., IEEE Trans. Nucl. Sci. NS-25 (1978) 1166. A W.L. Bendel and E.L. Petersen, IEEE Trans. Nud. Sci. NS-30 (1983) 4481.