Available online at www.sciencedirect.com

www.elsevier.com/locate/asr

Advances in Space Research 42 (2008) 1293–1299

A new modeling method of solar energetic proton eventsfor ISO specification

Yoichi Kazama *, Tateo Goka

Institute of Aerospace Technology, Japan Aerospace and Exploration Agency, Sengen 2-1-1, Tsukuba, Ibaraki 305-8505, Japan

Received 7 November 2006; received in revised form 21 December 2007; accepted 21 December 2007

Abstract

Solar energetic protons degrade performance and reliability of spacecraft systems due to single-event effects, total dose effects anddisplacement damage in electronics components including solar cells. On designing a solar cell panel, a total fluence of solar energeticprotons (SEPs) which cause solar cell damage is needed to estimate power loss of the solar cells over mission life. Nowadays a solar panelarea of spacecraft is increasing as spacecraft mission life becomes longer (15–18 years). Thus an accurate SEP model is strongly requiredfor the cost-minimum design from the aerospace industry. The SEP model, JPL-91 proposed by Feynman et al., is currently used widelyfor solar cell designing. However, it is known that the JPL-91 model predicts higher fluences of protons than values actually experiencedin space, especially after 7 years on orbit. In addition, the model is based on several assumptions, and also needs Monte-Carlo simula-tions for calculating fluences. In this study, we propose a new method for modeling SEPs especially focused on solar cell degradation.The newly-proposed method is empirical, which constructs a model based directly upon proton flux measurement data taken by instru-ments onboard spacecraft. This method has neither assumptions nor dependence on SEP event selection, both of which are needed inJPL-91. The model fluences derived from this method show lower fluences in longer missions compared to JPL-91. This method has beenproposed to ISO (International Organization for Standardization) and has been discussed for a new standard SEP model.� 2008 Published by Elsevier Ltd on behalf of COSPAR.

Keywords: Solar energetic proton; Spacecraft engineering; Solar cell degradation; ISO standard

1. Introduction

It is well known that solar energetic protons (SEPs)damage spacecraft systems, that is, electronics and solarcells due to ionization and/or atomic displacement pro-cesses. This results in single-event upsets and latch-ups inelectronics, and output deterioration of solar cells.

Solar cells of spacecraft are obviously one of the keycomponents of spacecraft systems. It is unavoidable thatenergetic protons degrade solar cells, which causes powerloss of the spacecraft systems. The degradation of cellsare crucial to its mission life. Therefore, an estimation ofSEP fluences in space is needed for designing solar cellpanels.

0273-1177/$34.00 � 2008 Published by Elsevier Ltd on behalf of COSPAR.

doi:10.1016/j.asr.2007.12.012

* Corresponding author. Tel.: +81 29 868 4213; fax: +81 29 868 5970.E-mail address: kazama.yoichi@jaxa.jp (Y. Kazama).

The most famous SEP model is JPL-91 proposed byFeynman et al. (1993), which has been widely used to esti-mate the fluences. The JPL-91 model is a probabilisticmodel, which means that the model assumes a probabilitydistribution of SEP events, and the parameters of the prob-ability distribution were determined by SEP measurementsby satellites during �2.6 solar cycles. The details of thismodel is described in the next section.

Recently, a new model of SEP events has proposed byXapsos et al. (2000). The model, which is referred to asthe ESP model, is also a probabilistic model similar tothe JPL-91 model, but the ESP model is based on the max-imum entropy theory, which is more reasonable comparedto the assumptions in JPL-91. Advantages of the ESPmodel are that (1) a wider energy range up to >300 MeVis covered, and (2) proton data in full three solar cycleswere used in constructing the model.

Fig. 1. Long-term cumulative proton fluence. The fluence was calculatedby using GOES satellite energetic proton data. JPL-91 fluences withseveral confidence levels are also shown by dotted lines for comparison.

1294 Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299

Another SEP event model has been presented by Nym-mik (1999) and was proposed to International Organiza-tion for Standardization, ISO, Technical Specification15391, 2004. This model is based on the assumption thatSEP events are correlated with sunspot numbers. Thismodel has the advantages that SEP fluences can be esti-mated even during quiet solar-activity periods in whichthe JPL-91 and ESP models assume zero fluences.Although these new models have been proposed, theJPL-91 model is still the de-facto standard model of SEPevents in the interplanetary space, and is widely used fordesigning spacecraft systems.

Solar cell engineers also use the JPL-91 model for estimat-ing solar cell degradation. However, in the point of view ofsolar cell degradation, it is well known that the JPL-91 modelpredicts higher SEP fluences than values actually experi-enced by spacecraft in space, especially after seven yearsfrom the launch. Nowadays, spacecraft manufacturers arevery conscious of cost-minimum design of spacecraftbecause lifetime of spacecraft becomes longer (15–18 years)and the cost of manufacturing spacecraft is increasing.Therefore, the aerospace industry requires a more accurateSEP fluence model for more realistic design of solar cells.

In the next section, we describe the widely-used SEPmodel JPL-91, and then explain a newly-proposed methodof modeling SEP events for solar cell engineers to estimateSEP fluences.

2. Currently-used JPL-91 model

In this section, the JPL-91 model is briefly reviewed toclarify the purpose of this study.

The JPL-91 model is constructed in the basis of a prob-abilistic treatment for SEP events, and gives a possible flu-ence which the spacecraft experiences in space for a givenmission life and a given confidence level. Here, a confidencelevel means a probability that the spacecraft does not expe-rience more than that fluence.

The model is constructed as follows: First, 11 years of asolar cycle are categorized into two periods: an active per-iod of 7 years and a quiet period of 4 years. At this stage, itis assumed that SEP events occur only in active periods anddoes not in quiet periods.

Second, SEP events are selected in a data set of energeticproton fluxes. The selection is made to satisfy that the totalfluence over an event is higher than a threshold level. Thenthe occurrence distribution versus event-integrated fluenceis made.

Under the assumption that the occurrence distributionobeys lognormal probability, the distribution is fitted to alognormal function. Finally, a Monte-Carlo simulation isdone to obtain probability–fluence curves for each missionlife of 1 year, 2 years, and so on. This calculation process ismade for several energy ranges of protons, that is, >1, >4,>10, >30 and >60 MeV.

Although this JPL-91 model is widely used for SEP flu-ence estimation, there are some points to which one should

be paid attention. For instance, in JPL-91, the occurrenceprobability of SEP events is assumed to a lognormal distri-bution. However, actual distributions do not always lookas lognormal distributions, (see the distributions, for exam-ple, in Feynman et al., 1993), especially for low-fluenceSEP events.

In addition, JPL-91 users obviously need to determinewhich confidence level they apply to their application. Usu-ally, the confidence levels are determined by other assump-tions and/or knowledge on SEP events.

Furthermore, JPL-91 needs Monte-Carlo simulations toobtain final probability–fluence curves, which is notfriendly to end users. It should be also noted that the modeldepends upon the criteria of event identification from theproton measurement data since SEP events often occurrecurrently.

To clarify these points, a long-term cumulative protonfluence at the geosynchronous orbit is shown in Fig. 1.Data used were 5-minute-averaged proton flux data mea-sured by GOES-5 to GOES-11 which cover from 1985through 2005, and were provided from Space Physics Inter-active Data Resource (SPIDR) at National Oceanic &Atmospheric Administration (NOAA) in the USA.

In the figure, the >4-MeV fluences in JPL-91 with sev-eral confidence levels are shown by dotted lines for com-parison. Solar active periods assumed in JPL-91 are alsoindicated in shaded areas.

As seen in the figure, proton fluences of 109/cm2 can beseen in the quiet period before 1989. JPL-91 assumes thatzero fluences exist during quiet periods, in which, howeverin fact, energetic protons exist.

The GOES cumulative proton fluence agrees with theJPL-91 fluences of 50–75% confidence levels in the longmission life. On designing a spacecraft, engineers assumea proton fluence with 90–95% confidence levels of JPL-

Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299 1295

91. Therefore, Taking today’s longer mission duration intoaccount, it is concluded that the widely-used JPL-91 modelgives higher fluences.

The tendency of this overestimation can also be seen inthe standard confidence levels of European Space Agency(ESA) in European Cooperation for Space Standardization(2001), which is tabulated in Table 1. ESA recommends dif-ferent confidence levels depending on the mission dura-tions: higher levels for shorter missions and lower levelsfor longer missions. This is due to the overestimation ofJPL-91, and they relax SEP fluence levels for longer mis-sions from the realistic point of view.

Resulting over-design of solar cells reflects a cost ofspacecraft, and a more realistic SEP model is requestedfrom the aerospace industry. In the present paper, a newmethod for modeling SEP fluences is proposed to try toanswer this need.

3. New modeling method

To describe a new modeling method, at first, we explainthe energy range of energetic protons important for solarcell degradation. Then, a proton data set used in this studyis described. Finally, model fluences of SEPs constructedby this method are shown.

3.1. Proton energy range

To see the energy range effective to solar cell degrada-tion, Fig. 2 shows energy profiles of Radiation DamageCoefficient (RDC) for protons taken from Assessment ofMultijunction Solar Cell Performance in Radiation Envi-ronment (2000). RDC indicates how much protons witha specific energy degrade the output of solar cells. HigherRDC means larger damage of solar cells. The panels givetwo types of three-junction solar cells, manufactured bySpectrolab (upper panel), and by TECSTAR (lower panel).

Looking at the profiles shown in red (solar cells with 3-mil-thick cover glass, widely-used for spacecraft), the pro-files rise up rapidly at �3 MeV. This is due to cutoff of pro-tons by the cover glass. The profiles increase as a protonenergy decreases since higher-energy particles have lessenergy deposition. In conclusion, proton energies from afew to 10 MeV are important to estimate solar cell radia-tion damage.

Table 1ESA’s standard confidence levels to be applied for various missiondurations

Exposure year Confidence level (%)

1 972 953 954 905 906 907 90

3.2. Proton data set

In the present study, we use proton flux data measuredby Space Environment Monitor (SEM) onboard three Jap-anese Geostationary Meteorological Satellites: GMS-2,GMS-3 and GMS-4 (Kohno, 1996). The data period isfrom December 21, 1981 to June 21, 1995.

Although the GMS satellites flew in the geosynchronousorbit, background energetic protons of the Earth’s radia-tion belts are negligible in the energy range of our interest,because the energetic proton model AP-8 (Sawyer andVette, 1976) indicates zero fluxes of protons near the geo-synchronous orbit, and consequently no proton fluxes ofradiation-belt origin can be seen in the geosynchronoussatellite data.

In this study, we focused on two energy channels of theSEM proton data, that is, 4–8 MeV (P2 channel) and 8–16 MeV (P3 channel) to cover the energy range importantfor solar cell radiation damage. As seen in Fig. 2, thesetwo channels well cover the energy range of interest to solarcell degradation.

Summary of the daily proton fluxes of GMS/SEM areplotted in Fig. 3. The P2 channel data are shown in the upperpanel, and P3 in the middle. Note that spurious noises wereremoved by visual inspection, and data points of very lowfluxes were also removed by applying the threshold of9/cm2 s sr for both 4–8 and 8–16 MeV channels.

For reference, sunspot numbers are also given in thebottom panel. The sunspot number data are taken fromthe on-line catalog of sunspot data of Solar InfluencesData Analysis Center (SIDC) at http://www.sidc.be/sun-spot-data/. It is seen that SEP events are well correlatedwith solar activities as indicated in the sunspot numbers.

Before using the data, we did cross-calibration of theSEM data with GOES data to check data consistency.Fig. 4 indicates the correlations of SEP events betweenGMS and GOES proton data. Proton fluxes of 4–8 and8–16 MeV are shown in the left and right panels, respec-tively. One can see clear correlations between two satellitedata in both the energy range. Thus, it is concluded that theGMS data quality is good. Note that low-flux events showdiscrepancies, but these events occupy only 0.05–0.08% ofthe total fluence over the whole period. Accordingly, thediscrepancies do not affect the calculation.

3.3. Calculations and results

In this section, the calculation method and modeledSEP fluences are described. While JPL-91 is based on aprobability consideration, we take a practical way to cal-culate model fluences by directly integrating long-termSEM/GMS proton flux data without assuming confidencelevels.

In our method, a model fluence for a given missionduration of n years is obtained as follows: (1) calculaten-year fluences by integrating daily fluences with shiftingthe integration window day-by-day, and (2) take the

Fig. 2. Radiation Damage Coefficients (RDCs) of protons taken from Assessment of Multijunction Solar Cell Performance in Radiation Environment(2000). RDCs of three-junction cells manufactured by Spectrolab and TECSTAR are shown in the upper and lower panels, respectively.

1296 Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299

maximum of the set of the integrated fluences as themodel fluence for the n-year mission duration. A sche-matic illustration for the steps is displayed in Fig. 5. Inother words, a model fluence given by the method corre-sponds to the maximum fluence if one launches a satel-lite everyday from the first day to the last of the dataperiod.

As a result, our model gives a possible fluence in theworst case. However, it should be noted that themethod does not take into account large SEP eventsnever happening during the data period. It is also tobe noted that since the proton data set used in thisstudy was taken from late 1981 to middle 1995, themodel fluences reflects the nature of SEP events seenin the 22nd solar cycle.

The results of the model fluences are shown in Figs. 6and 7 for 4–8 and 8–16 MeV protons, respectively. Boththe figures have also JPL-91 fluences with confidence levelsof 50%, 75%, 90%, 95% and 99% (dotted lines). The energyrange of the JPL-91 fluences are >4-MeV for the 4–8-MeV

result, >10 MeV for the 8–16-MeV result. Note that weassume first 7 years are a solar active period to plot theJPL-91 fluences.

In the first few years of a mission duration, the modelfluences show agreement with the JPL-91 fluences withconfidence levels more than 90% or 95%. This comes fromhigh fluxes in active periods since we took the maximum offluences.

As the mission duration becomes longer, the confidencelevel corresponding to the model fluences become lower.Finally, the model fluences go to a 50%–75% level inthe 4–8-MeV proton case, and a 75%–90% level in the8–16-MeV proton case. These final confidence levels aresmaller than 90%–95% which are taken into account indesigning solar cells. Considering that mission life of sat-ellites nowadays becomes longer (15–18 years), to takeconfidence levels 90%–95% of JPL-91 is overestimationof SEP fluences.

Since the purpose of this study is focused especially onestimation of solar cell deterioration caused by SEP flu-

Fig. 3. Summary plot of daily proton fluences taken by GMS-2/3/4. The upper and middle panels show fluences of 4–8 and 8–16 MeV protons,respectively. Sun spot numbers are given in the bottom panel.

Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299 1297

ences in space, we propose an empirical and straightfor-ward method of modeling SEP fluences, consideringadvantages brought from its simpleness: (1) the methodgives realistic fluences of SEPs in space, (2) the methoddoes not need to set confidence levels, (3) neither event

identification nor assumptions of event probability distri-butions is needed, and (4) it is easy for end users to usethe method.

In this paper, we use proton flux data of 1.2 solar cycle,which is shorter than that used in JPL-91. It is, however,

Fig. 4. Correlations of SEP events between the GOES and GMS data. The left and right panels show 4–8-MeV proton fluxes and 8–16-MeV proton fluxes,respectively. Clear correlations between GOES and GMS are seen in both the panels.

data set of daily fluxes

A

B

C

Z

time

Sun Spot #

F(n) = max{Σ(A), Σ(B), Σ(C), ..., Σ(Z)}, n = 1, 2, 3, ... [year]

1 day

2 days

n

0 day

Fig. 5. Schematic illustration to explain how to calculate a model fluencein our method. Cumulative fluences are calculated by integrating the datawith shifting the integration window, and then the maximum of thefluences is taken as the model fluence for the given duration.

Fig. 6. Model fluence with energies of 4–8 MeV. The dotted lines show>4-MeV fluences of JPL-91 with confidence levels of 50%, 75%, 90%, 95%and 99%. In longer mission durations, the model fluence corresponds to aconfidence level of 50–75%, which are not as high as are currently takeninto account when designing solar cells.

1298 Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299

emphasized that the essence of this study is to show a newmodeling method of SEP events. This method is clearlyapplicable for other data sets, and investigating the differ-ences of the results between different data sets is a futurework. It will be also discussed which data set is to be usedfor constructing a standard SEP model.

4. Summary

We have proposed a new method of modeling solarenergetic proton (SEP) fluences for engineering applica-tions, especially for estimation of solar cell degradation.

By using this method, the model fluences of SEPevents are shown based on the long-term proton fluxdata of Japanese Geostationary Meteorological Satel-lites (GMS).

The method takes a pragmatic approach, that is, thetotal fluences are calculated directly from long-term ener-getic proton data measured in space, and the method hasseveral merits such as: (1) more realistic SEP fluences, (2)no confidence level needed, (3) neither event identificationnor assumptions needed, and (4) user-friendliness.

We have proposed this modeling method as a newSEP model for solar cell designing to ISO (InternationalOrganization for Standardization), technical committee20 (TC20) ‘Aircraft and space vehicles’, sub-committee14 (SC14) ‘Space systems and operations’, working group

Fig. 7. Model fluences with energies of 8–16 MeV. The dotted lines show>10-MeV fluences of JPL-91 with confidence levels of 50%, 75%, 90%,95% and 99%. In longer mission durations, the model fluence correspondsto a confidence level of 75–90%.

Y. Kazama, T. Goka / Advances in Space Research 42 (2008) 1293–1299 1299

4 (WG4) ‘Space environment (natural and artificial)’.The model has been selected as an ISO Technical Spec-ification, and has been discussed in the ISO workinggroup.

Acknowledgement

We thank Space Physics Interactive Data Resource(SPIDR) at National Oceanic and Atmospheric Adminis-tration (NOAA) for providing GOES proton data.

References

Assessment of Multijunction Solar Cell Performance in RadiationEnvironment, The Aerospace Corporation, February 29, 2000.

Feynman, J., Spitale, G., Wang, J., Gabriel, S. Interplanetary protonfluence model: JPL 1991. J. Geophys. Res. 98, 13281–13294, 1993.

Kohno, T. Current and future data available in Japan. Geophys. Monogr.Ser. 97, 217–222, 1996.

Nymmik, R.A. Probabilistic model for fluences and peak fluxes of solarenergetic particles. Radiat. Meas. 30, 287–296, 1999.

Sawyer, D.M., Vette, J.I. AP8 Trapped Proton Environment For SolarMaximum and Solar Minimum, National Space Science Data Center,Report 76-06, Greenbelt, Maryland, 1976.

Space Engineering, Space Environment, European Cooperation for SpaceStandardization (ECSS-E-10-04A), ESA-ESTEC, Requirements &Standards Division, Noordwijk, The Netherlands, 2001.

Space Environment (Neutral and Artificial), Probabilistic model forfluences and peak fluxes of solar energetic particles. Part I. Protons,International Organization for Standardization (ISO), TechnicalSpecification 15391, 2004.

Xapsos, M.A., Summers, G.P., Barth, J.L., Stassinopoulos, E.G., Burke,E.A. Probability model for cumulative solar proton event fluences.IEEE Trans. Nucl. Sci. 47, 486–490, 2000.