1. © Copyright QinetiQ limited 2006 Exploitation of GIOVE in-Orbit Radiation Data for Environment Model and Effects Tools Update ESA Contract Number: 4000105611 Final Presentation A. Hands, B. Taylor, K. Ryden & C. Underwood

2. © Copyright QinetiQ limited 2006 2 16 th December 2015, ESTEC Background The Galileo constellation will operate in an intense radiation environment (Medium Earth Orbit) The outer Van Allen electron belt poses significant radiation risks due to internal charging and ionising dose Other threats exist from Solar Particle Events Existing models for worst case environment definition are out-of- date New data provide an opportunity to re-examine Galileo’s radiation environment specification Galileo In-Orbit Validation Element (GIOVE) spacecraft equipped with monitors to address these issues

3. © Copyright QinetiQ limited 2006 3 16 th December 2015, ESTEC Giove Spacecraft Technology Demonstrator Satellites for Galileo Constellation Each Carries Space Environment Monitor(s) Medium Earth Orbit (~23,500 km, 56°) Giove-B: Launched Apr 2008 One instrument: SREM Giove-A: Launched Dec 2005 Two instruments: Merlin Cedex

4. © Copyright QinetiQ limited 2006 4 16 th December 2015, ESTEC Proposal Objectives To create a calibrated, background-subtracted, clean, electron flux dataset derived from Cedex and Merlin and make comparisons with equivalent SREM data. To cross-calibrate the proton and ion LET spectra from Cedex, Merlin and SREM. To validate existing radiation belt models against the data from Giove-A and –B sensors and assess their accuracy. To assess the Galileo FOC satellite radiation specification against the new data and recent models. To create a new internal charging-oriented radiation belt model as an update to the existing FLUMIC model. To exploit the single event upset (SEU), dose, dose-rate and charging current information from Giove instruments and model these effects.

5. © Copyright QinetiQ limited 2006 5 16 th December 2015, ESTEC Project Overview Duration (initially) 18 months Led by QinetiQ, Surrey Space Centre subcontractor Various staffing changes led to SSC taking over full project Split into five work packages: Project Plan Novation to SSC

6. © Copyright QinetiQ limited 2006 6 16 th December 2015, ESTEC Cedex Photodiodes: Sensitive to dose-rate Four diodes with different shielding levels Merlin SURF monitor: Current collecting plates (direct measure of internal charging) Electrons only SREM Silicon Detectors: Telescopic arrangement Deconvolution algorithm to separate electrons and protons Task 1: Electron calibration and data processing Data coverage periods: Cedex & Merlin (Giove-A): December 2005 – July 2012 SREM (Giove-B): April 2008 – July 2012 Electron-detecting Instruments:

7. © Copyright QinetiQ limited 2006 7 16 th December 2015, ESTEC Merlin-SURF SURF Three stacked aluminium charge-collecting plates Direct measure of energetic electrons No proton contamination or dead-time Targeted at internal charging concerns

8. © Copyright QinetiQ limited 2006 8 16 th December 2015, ESTEC SURF Data First Day: First 6 months: Recurrence of coronal hole Samples peak of belt every ~7 hours

9. © Copyright QinetiQ limited 2006 9 16 th December 2015, ESTEC SURF Instrument Response Functions Response functions determined by Monte Carlo simulations Electrons from <1 MeV to ~10 MeV detected by SURF Lower plate has strong response due to increased thickness ‘Free’ from proton contamination (v. small opposite polarity currents during SPE) E min ~ 0.5 MeV

10. © Copyright QinetiQ limited 2006 10 16 th December 2015, ESTEC Proton Contamination No increases in SURF currents during 2012 SPEs For completeness, proton response functions also calculated Predicted +ve currents (fA/cm 2 ) during major SPE: cf. electron (-ve) currents (pA/cm 2 ) I.e. even worst case SPE +ve currents << | -ve currents | ￫ No proton contamination (very few SPE during project period in any case)

11. © Copyright QinetiQ limited 2006 11 16 th December 2015, ESTEC Electron Flux Determination Need flux, not current, for inter-instrument comparison and model development Simple iterative algorithm developed: Best fit algorithm Exponential or Power-law spectrum Observed currents Convolve with SURF plate response functions Adjust folding energy / spectral index and amplitude For more detail see Ryden et al. (2008) IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 36, NO. 5, OCTOBER 2008

12. © Copyright QinetiQ limited 2006 12 16 th December 2015, ESTEC Example of Electron flux (assuming exponential spectrum) Measure of spectral shape (‘hardness’) NB Flux most reliable in range: 0.5 – 3 MeV

13. © Copyright QinetiQ limited 2006 13 16 th December 2015, ESTEC Cedex Geant4 modelling of dose-rate diode response functions Forward and reverse Monte Carlo techniques used

14. © Copyright QinetiQ limited 2006 14 16 th December 2015, ESTEC Cedex off-set correction Dose-rate diodes experience variable noise outside radiation belt: Used to clean Cedex data

15. © Copyright QinetiQ limited 2006 15 16 th December 2015, ESTEC Electron spectra determination Ratio between dose-rate diodes used to derive exponential electron flux spectra: Requires correlation between diode dose-rates: No correlation (unused) Compare with calculated ratio Determine spectral folding energy

16. © Copyright QinetiQ limited 2006 16 16 th December 2015, ESTEC SREM Data Derivation of flux not part of this study SREM fluxes calculated by NOA using SVD analysis of 15 SREM channels (proton + electron) E.g.: Minimal contamination outside Van Allen belt Electron Flux

17. © Copyright QinetiQ limited 2006 17 16 th December 2015, ESTEC Merlin-SURF vs. Cedex Good differential flux agreement at 1 MeV and 2 MeV: 2 MeV 1 MeV Shielded Cedex diodes only useable in high flux periods

18. © Copyright QinetiQ limited 2006 18 16 th December 2015, ESTEC Merlin-SURF vs. Cedex Better agreement with SURF spectra during high flux periods: All flux High flux only → Cedex deemed less reliable in low flux periods

19. © Copyright QinetiQ limited 2006 19 16 th December 2015, ESTEC Merlin-SURF vs. SREM Good agreement, especially during high flux periods (equatorial crossings where L~5):

20. © Copyright QinetiQ limited 2006 20 16 th December 2015, ESTEC Merlin-SURF and SREM electron spectra cf. AE8 Spectral comparison (equatorial): All instruments show harder spectrum than AE8 (but low confidence due to response function range) Better agreement with AE8 with longer SURF data coverage period

21. © Copyright QinetiQ limited 2006 21 16 th December 2015, ESTEC Cedex Two Silicon Detectors: Telescopic arrangement LET range ~ 30 – 15,000 MeV.cm 2 /g SREM Three Silicon Detectors: Telescopic arrangement Deconvolution algorithm to separate electrons and protons Merlin Particle telescopes: Single channel (>40 MeV) proton telescope) Ion LET telescope – no data due to hardware defect Task 2: Proton and Heavy Ion calibration and data processing Data coverage periods: Cedex & Merlin (Giove-A): December 2005 – July 2012 SREM (Giove-B): April 2008 – July 2012 Proton / Heavy ion instruments:

22. © Copyright QinetiQ limited 2006 22 16 th December 2015, ESTEC Merlin Proton Telescope Stacked diodes – diameter ~2 cm Detects protons >40 MeV Also detects electrons due to pile-up: Protons (>40 MeV flux) Electron contamination when bottom SURF plate ≳0.02 pA/cm 2 (no real proton signal in this period)

23. © Copyright QinetiQ limited 2006 23 16 th December 2015, ESTEC Solar Particle Events Several SPEs measured (largest in 2012) Dips in measured proton flux due to geomagnetic shielding (correlate with equatorial peaks in electron intensity)

24. © Copyright QinetiQ limited 2006 24 16 th December 2015, ESTEC May 2012 SPE Smaller event in May shows complex structure of proton flux with electron contamination Proton peak Electron contamination Dip (geomagnetic shielding) Electron contamination within dip SURF Bottom Plate

25. © Copyright QinetiQ limited 2006 25 16 th December 2015, ESTEC SREM Proton Flux As with electron flux, need to de-convolve electron & proton contributions (done by NOA) Electron contamination evident at equatorial crossings: Proton Flux Electron contamination Need to restrict proton analysis to periods outside electron belt (e.g. L>10)

26. © Copyright QinetiQ limited 2006 26 16 th December 2015, ESTEC 2012 SPEs >40 MeV proton flux calculated from SREM data for comparison to Merlin E.g. January 2012: Reasonable agreement at peak flux Merlin < SREM at low flux

27. © Copyright QinetiQ limited 2006 27 16 th December 2015, ESTEC 2012 SPEs March 2012 May 2012

28. © Copyright QinetiQ limited 2006 28 16 th December 2015, ESTEC Cedex LET Telescope Two 3x3 cm, ~300 um thick PIN silicon detectors (under 2.5 mm copper dome) No electron contamination: Large increase in electron- induced dose does not result in significant increase in integral ion flux Dose Rate: Ion flux:

29. © Copyright QinetiQ limited 2006 29 16 th December 2015, ESTEC LET Spectra Counts binned into 512 linearly spaced LET channels LET range: ~ 30 – 15,000 MeV.cm 2 /g Threshold used to exclude SPEs

30. © Copyright QinetiQ limited 2006 30 16 th December 2015, ESTEC LET Spectra - Comparison with Models Quiet time: Including SPEs: Reasonable agreement with CREME model

31. © Copyright QinetiQ limited 2006 31 16 th December 2015, ESTEC Long term variation Inverse correlation between ion flux and solar cycle due to modulation of Galactic Cosmic Rays (GCR)

32. © Copyright QinetiQ limited 2006 32 16 th December 2015, ESTEC Ion Flux during SPEs 2012 events measured Dips during equatorial crossings visible: January 2012 March 2012 May 2012

33. © Copyright QinetiQ limited 2006 33 16 th December 2015, ESTEC Task 3: Radiation Belt Model Validation Objectives: 1.Analyse Giove electron flux data and compare to models (e.g. AE8, MEO, FLUMIC, AE9) in order to identify deficiencies or inconsistencies 2.Examine data with respect to Galileo radiation environment specification - assess suitability for the mission Data sets: Electron fluxes derived from Merlin-SURF (2005 – 2012) - exponential and power-law spectra treated separately Electron fluxes derived from SREM (2008 – 2012) (Cedex-derived fluxes excluded as only available during high flux periods and low fidelity) Focus on SURF data (longest data set and best traceability)

34. © Copyright QinetiQ limited 2006 34 16 th December 2015, ESTEC AE8 AE8 (still) represents industry standard Static model, no flux variability Inner and outer belts (only latter applicable to Giove) Giove minimum L

35. © Copyright QinetiQ limited 2006 35 16 th December 2015, ESTEC Time Series Comparison SREM differential flux cf. AE8: 0.7 MeV 1.4 MeV Big discrepancy during ‘electron desert’ period Reasonable agreement 2010 – 2012 period AE8 ‘confidence levels’

36. © Copyright QinetiQ limited 2006 36 16 th December 2015, ESTEC Time Series Comparison SURF differential flux cf. AE8: Exponential Spectrum Power-law Spectrum 0.5 MeV 1.5 MeV Longer SURF time period highlights unusual nature of electron desert

37. © Copyright QinetiQ limited 2006 37 16 th December 2015, ESTEC Spectral Comparison – AE8 Mean long-term spectra derived from SREM and SURF data (excluding L>8 data) AE8 spectrum averaged over Galileo-type orbit SREM: SREM < AE8 at low energies (and harder spectrum)

38. © Copyright QinetiQ limited 2006 38 16 th December 2015, ESTEC Spectral Comparison – AE8 SURF spectra for exponential and power-law assumption (extended beyond optimum 0.5 – 3 MeV range for illustration) SURF exponential spectrum < AE8 at low energies SURF power-law spectrum >AE8 at low energies Both >AE8 at high energies – danger of extrapolation!

39. © Copyright QinetiQ limited 2006 39 16 th December 2015, ESTEC Spectral Comparison - FLUMIC FLUMIC is a worst case daily-averaged electron flux model for internal charging Comparison with representative low, medium and high flux periods for SREM and SURF fluxes (for statistical comparison see Task 4) FLUMIC parameters: L=4.5, equatorial latitude, two representative dates (determines normalisation and folding energy) ‘FLUMIC 1’ = September 2007, E 0 = 0.48 MeV, ‘FLUMIC 2’ = January 2001, E 0 = 0.28 MeV Giove data imply FLUMIC is too soft (electron spectrum actually hardens during enhancements) FLUMIC specification is exceeded at high energies (margin depends on extrapolation assumptions)

40. © Copyright QinetiQ limited 2006 40 16 th December 2015, ESTEC Spectral Parameters vs. Intensity Relationship between spectral hardness and intensity of electron environment (applies to flux as a function of magnetic latitude and variation due to solar wind conditions) Folding energy increases with intensity (harder spectrum) Spectral index decreases with intensity (harder spectrum) SURF (exponential spectrum) SURF (power–law spectrum)

41. © Copyright QinetiQ limited 2006 41 16 th December 2015, ESTEC Galileo Environment Specification Radiation environment specification addresses (primarily) concerns over ionising dose and internal charging Specification w.r.t. models: Specification based on AE8 AE9 < AE8 in Galileo orbit (except at high E)

42. © Copyright QinetiQ limited 2006 42 16 th December 2015, ESTEC Galileo Environment Specification With respect to Giove data: At 3 MeV only SURF power–law is above specification If extrapolated to higher energies, all Giove spectra exceed Galileo specification

43. © Copyright QinetiQ limited 2006 43 16 th December 2015, ESTEC Dose-Depth However, this is highly sensitive to upper energy cut-off… Varying spectra have significant consequences for dose-depth specification: Upper energy cut-off for these examples: SREM: 3.8 MeV SURF: 5.5 MeV (Dose for nominal 12 year period)

44. © Copyright QinetiQ limited 2006 44 16 th December 2015, ESTEC Dose-Depth – dependence on upper cut-off energy Illustrative calculation for different (extrapolated) upper energy cut-offs: Significant difference with exponential spectrum at ~8 mm shielding (but much less significant than power-law extrapolation) Very large difference with power-law spectrum above ~6 mm shielding (extrapolation unjustified?) SURF upper cut- off varied from 3.5 - 10 MeV

45. © Copyright QinetiQ limited 2006 45 16 th December 2015, ESTEC Dose-Depth What to conclude for Galileo environment specification? Cannot justify simplistic extrapolation of spectra beyond ~3 MeV However, also cannot justify truncating spectrum at 3 MeV for dose-depth calculation Reasonable compromise: create composite dose-depth curve based on AE8 and ONERA’s Galileo specification model (based on GPS data but better agreement with Giove data than AE8) Consequence for Galileo specification: ~doubling of dose at 8 mm shielding (unchanged 12 mm)

46. © Copyright QinetiQ limited 2006 46 16 th December 2015, ESTEC Task 4: Model of Outer Belt Electrons for Dielectric Internal Charging (MOBE-DIC) Energetic trapped electrons in Van Allen belts pose a threat to satellites through internal charging of dielectric materials: The outer electron belt is extremely dynamic - large changes in flux occur over short timescales, driven by coronal holes and coronal mass ejections (CMEs) Existing models to address internal charging: FLUMIC: Worst-case model for internal charging Based primarily on GEO data (not near peak) User-friendly but not up-to-date AE9: Successor to AE8 Multiple data sources Comprehensive statistics Complex (many input parameters & run options) Objective: A fast user-friendly model for internal charging specification (successor to FLUMIC)

47. © Copyright QinetiQ limited 2006 47 16 th December 2015, ESTEC Internal Charging e.g. April 2010: >2 MeV flux at GEO increases by ~4 orders of magnitude in a few hours! Electrostatic charging of spacecraft materials Electrostatic discharge Energy coupling into circuit Satellite anomaly / outage /failure Multi-stage process:

48. © Copyright QinetiQ limited 2006 48 16 th December 2015, ESTEC Worst Case Statistics with Giove Data Use derived flux time series to create cumulative distribution functions (CDFs) at discrete energies in the range 0.5 – 3 MeV (peak of instrument response) (Note: harder spectra in more extreme events) Fit baseline spectra at three percentage exceedance levels - 90%, 99% & 100%: Create reduced series of average flux at equatorial peaks (L≈4.7) CDFs based on equatorial flux: These three spectra form the basis of the MOBE-DIC model (more in Hands et al. IEEE TNS 2015) NB SURF data used (2005 – 2012)

49. © Copyright QinetiQ limited 2006 49 16 th December 2015, ESTEC Extrapolating to other L-Shells Equatorial spectra at L≈4.7 form the basis of the model Need to derive profile of L-shell to extrapolate, however… L-Shell profile is not stable, e.g.: SURF on STRV1d (Ryden et al. 2001) Van Allen Probes, REPT (Baker et al. 2014) AE8 AE9 FLUMIC Van Allen Probes, ERM (Maurer et al. 2013)

50. © Copyright QinetiQ limited 2006 50 16 th December 2015, ESTEC Extrapolating to other L-Shells (2) Our approach is to use high-latitude SURF data Inclination of Giove-A orbit means higher L shells only encountered at higher latitudes Need to renormalise non-equatorial fluxes: L≈4.7 at equator Assume Vette function (like AE8 and FLUMIC) [Scaling is (slightly) L-dependent but not energy-dependent] Equatorial Flux All Flux Fit ‘envelope’ to renormalised data (at each energy) ￫ Energy-dependent L-Shell profile

51. © Copyright QinetiQ limited 2006 51 16 th December 2015, ESTEC Extrapolating to other L-Shells (3) Final (energy-dependent) L-shell profile (3 < L < 8) FLUMIC function used below L=4.5 (no Giove data) Normalised to L=4.7: Normalised to L=6.6: (NB slightly modified version used for integral flux)

52. © Copyright QinetiQ limited 2006 52 16 th December 2015, ESTEC Comparison to GOES data Equivalent CDFs constructed from GOES (geostationary) electron flux data Compare with MOBE-DIC prediction at L=6.6: >2 MeV flux adjusted to L=6.6 and for dead-time effects (Meredith et al., 2015) MOBE-DIC prediction for ‘100%’ (worst case) at GEO for >2 MeV flux is: 2.34 x 10 5 e/cm 2 /s/sr Theoretical upper limit (Koons et al. 2001)… 2.34 x 10 5 e/cm 2 /s/sr !! Good agreement between MOBE- DIC and GOES at 99% and 100% (slightly worse at 90% due to conservative L-shell envelope) Unsurprisingly, excellent consistency at L=4.7 (MEO)

53. © Copyright QinetiQ limited 2006 53 16 th December 2015, ESTEC Comparison to existing models Comparison to FLUMIC model (including ALE) : Integral Spectra MOBE-DIC gives harder spectrum at MEO (but similar to FLUMIC ALE) MOBE-DIC exceeds NASA HDBK internal charging specification at 100% level

54. © Copyright QinetiQ limited 2006 54 16 th December 2015, ESTEC Comparison to existing models AE9: Differential Spectra (MOBE-DIC banded due to L-shell variation in AE9 orbits) 4.7< L< 5.1 6.6< L< 7.4 Longitude effect MOBE-DIC has harder spectrum than AE9 at MEO

55. © Copyright QinetiQ limited 2006 55 16 th December 2015, ESTEC MOBE-DIC: Implementation MOBE-DIC model is defined by a set of parameters and simple equations At present simple spreadsheet implementation: Public version available on request (a.hands@surrey.ac.uk) To be made available via Spenvis…

56. © Copyright QinetiQ limited 2006 56 16 th December 2015, ESTEC Task 5: Effects Characterisation Overview: Giove instruments directly measure radiation effects: Dose-depth 2 x RadFETs on Merlin 4 x Dose-rate photodiodes on Cedex Internal charging Three charge collecting plates on Merlin Single Event Effects Single event upsets (SEU) recorded by Giove-A on-board computer (OBC) Objective: use these data to compare measured effects to model predictions

57. © Copyright QinetiQ limited 2006 57 16 th December 2015, ESTEC Dose-Depth Measurements Merlin RadFETs NB Dose increases dominated by electrons (rather than SPE) 6 mm shield 3 mm shield Evidence of fade Fade corrected

58. © Copyright QinetiQ limited 2006 58 16 th December 2015, ESTEC Dose-Depth Measurements 4 x Cedex photodiodes (corrected for noise using signal outside Van Allen belt): Behind 4 mm Al Behind 2 mm Al Behind 2 mm CuBehind 4 mm Cu Correction applied

59. © Copyright QinetiQ limited 2006 59 16 th December 2015, ESTEC Dose-Depth vs Models Comparison with Shieldose predictions based on AE8 and SURF spectra: With 3 MeV cut-off, prediction << data at 6 mm shielding Models > Data at low shielding (but significant uncertainty over data correction) Circles = Cedex photodiodes Triangles = Merlin RadFETs Models and data converge at high shielding levels

60. © Copyright QinetiQ limited 2006 60 16 th December 2015, ESTEC Charging Current vs. Depth Monte Carlo Simulations used to create simple function of charging current as a function of shielding depth for MOBE-DIC spectra – generalised to all exponential spectra: Exponential transmission of current through shielding: Power-law relationship between d 0 and E 0 : General form: Discrete MOBE-DIC spectral values

61. © Copyright QinetiQ limited 2006 61 16 th December 2015, ESTEC Single Event Effects Giove-A OBC record of single event upsets (SEU): Background SEU rate caused by Galactic Cosmic Rays (varies inversely with solar cycle)

62. © Copyright QinetiQ limited 2006 62 16 th December 2015, ESTEC SEU rates during SPEs 2006 2012 Events that don’t lead to SEUs: (soft proton spectra) Events that do lead to SEUs: (hard proton spectra)

63. © Copyright QinetiQ limited 2006 63 16 th December 2015, ESTEC SEU Rate Calculations What is the balance between proton-induced SEU and ion-induced SEU? Need (but don’t have) ground test data for specific memory (Samsung SRAM) Use proxy data for ‘similar’ parts – “Rev B” and “Rev C” “Rev B”: (QinetiQ proton/neutron tests) “Rev C”: (Hirex proton tests) ~factor 30 range in SEU proton cross- section σ SEU ≈ 6 x 10 -14 cm 2 σ SEU ≈ 2 x 10 -15 cm 2

64. © Copyright QinetiQ limited 2006 64 16 th December 2015, ESTEC Predicted proton-induced SEUs cf. OBC Data SEUs during SPEs provide constraints on proton SEU cross-section: OBC measured rates GCR Predicted rates < measured rates (Rev B and Rev C) SPE: Predicted rates < measured rates (Rev C) Predicted rates >> measured rates (Rev B) Proton data imply “Rev C” SRAM is closer approximation Consistent with GCR SEU rate dominated by ions (LET) (IRPP calculations performed but subject to large uncertainties)

65. © Copyright QinetiQ limited 2006 65 16 th December 2015, ESTEC Future Impending missions will bring in more data from similar instruments… GalileoDSX Two FOC spacecraft in constellation will carry EMU monitor (successor to Merlin with 8 SURF charging plates) Launch 2016? AFRL/LWS mission with SET-1 payload including CREDANCE monitor (identical to Merlin) Slot region orbit: ~6,000 x 12,000 km Launch September 2016?

66. © Copyright QinetiQ limited 2006 66 16 th December 2015, ESTEC Project Results Summary Task 1: Electron calibration and data processing Flux series derived and cross-calibrated Task 2: LET spectra and Proton analysis Proton events and LET spectra analysed Task 4: FLUMIC model successor MOBE-DIC model created and validated Task 5: Effects characterisation SEUs, Dose-Depth and Charging currents investigated Task 3: Radiation belt model validation Giove data compared to models and Galileo FOC specification