Geant4 and the Vanderbilt Radiation Effects Simulation Strategy, RADSAFE

Transcription

1 Geant4 and the Vanderbilt Radiation Effects Simulation Strategy, RADSAFE Robert A. Weller, Robert A. Reed & Marcus H. Mendenhall Institute for Space & Defense Electronics Vanderbilt University Additional Collatorators: K. M. Warren D. R. Ball J. A. Pellish B. D. Sierawski C. L. Howe A. D. Tipton R. D. Schrimpf L. W. Massengill M. Alles L. Sternberg A. F. Witulski B. E. Templeton NASA Goddard: M. A. Xapsos K. A. LaBel NASA Marshall: J. H. Adams J. W. Watts Sponsoring Agencies: NASA, DTRA, AFOSR, AEDC Geant4 Space Users Workshop - Pasadena - 7/Nov/2006 1

2 Overview Objective - A Priori radiation effects simulation in semiconductors Motivation - Current methods are increasingly encountering problems Strategy - RADSAFE Geant4 application - MRED Details - What s in MRED and why Applications - The next talk, R. Reed 2

3 RADSAFE The Vision Predictable System Error Rates Trackable System Effects Statistical Circuit Response Functional Errors 3 Component Response Statistics Physics of Radiation Interaction

4 The RADSAFE System A system for first-principles predication of device, circuit and system response to radiation. Event-based Analysis RADSAFE is a strategy for implementing the third paradigm of scientific discovery, high fidelity computer simulation, in the field of radiation effects in semiconductor electronics and materials. 4

5 Structure of MRED MRED8 is the first generation Python/Geant4 application Target machine is a Linux cluster with!1k x86 and ppc nodes Development is with Xcode under Darwin (Mac OS 10.4) Presently using gcc 4 Python SWIG MRED C++ Geant4 5

6 MRED in RADSAFE 6

7 Python control shell MRED8 Features Standard internal data structure for radiation environments TCAD structure file parsing Constructive solid tetrahedron Now G4Tet! General voxel array input mechanism Screened Coulomb scattering and recoil tracking Custom particle gun derived from G4ParticleGun Three-level track weighting, ion selection, energy selection, hadronic interaction biasing Interpolating function or C2Function tool Detector class with multiple sensitive volumes Plug-in socket for combining sensitive volume energies HDF5 output files A WWW interface for remote control 7

8 MRED8-TCAD Interface Protons incident on an advanced CMOS integrated circuit Reactions in the metal layers increase energy deposition Proton Beam Radiation Events Device Description 8

9 MRED8 Voxel Array Processor Supplements TCAD inputs Detectors and regions defined in the input file Color by region name Supports extended applications such as medical dosimetry Image: M.H.Mendenhall 9

10 MRED8 Physics Run-time selection of custom MRED physics or any Geant4 physics list Coulomb scattering of nuclei and tracking of recoils Hadronic cross section biasing with automatic track weighting Custom particle gun Derived from G4ParticleGun class Supports random planar and isotropic flux simulation Uses C2Functions class to implement random energy from arbitrary input distributions Random energy selection from the integral distribution -oruniform or logarithmic selection and track weighting Uniform track-weight scaling for radiation environments 10

11 MRED8 Displacement Single Event 11

12 Running a Space Environment Si-Nitride 0.4!m SiO2 1.0!m Al 0.84!m SiO2 0.60!m Al 0.45!m SiO2 or W 0.6!m Al 0.45!m SiO2 0.6!m Si 0.25!m 2x2x2!m 3 Sensitive Volume 50!m 12

13 Effects of Heavy Elements Si-Nitride 0.4!m Direct SiO2 1.0!m Al 0.84!m SiO2 0.60!m Al 0.45!m SiO2 or W 0.6!m Al 0.45!m SiO2 0.6!m Si 0.25!m 50!m 2x2x2!m 3 Sensitive Volume 13

14 MRED8 Detectors & Event Selection MRED8 is basically a calorimetry tool Detector class includes rectangular and ellipsoidal sensitive volumes in arbitrary numbers. Each SV includes a validity range and invert logic flag. Collected charge model - Linear or arbitrary combinations of sensitive volume energies Hard wired filter logic implements first-level event selection to: Include nuclear reaction events Include Coulomb elastic scattering events Include hadronic elastic scattering events Include delta-ray event Include large-angle scattering events by the projectile Include events satisfying energy-based validity criteria Do boolean logic on sensitive volume and detector validity Custom histogram class with Python analog class General, programmable hit mechanism for particle/location/direction selection 14

15 Sensitive Volume Descriptions Defined as rectangular volumes Can be nested Linear combinations model charge collection Volumes represent diffusion regions of off-state transistors, High efficiency, small areas Arbitrary functions possible using plug-ins Volumes below the STI, low efficiencies, large areas 15

16 MRED8 Available Outputs Primary output based on HDF5 data model MRED8 can pass full G4Event objects to Python Sensitive volume histograms A validity map - a histogram implemented as an STL map indexed by a key constructed from the validity bools of all SVs Mathematica output for backward compatibility 16

17 HDF5 File Example 17

18 Writing the HDF5 Tables 18

19 Conclusion The heavy fragment distributions from ion-ion collisions must be modeled for SEE Fe on anything is essential Xe on anything is almost as important Super-heavy intermediate nuclei must not crash Geant4 Good physics is important; limits are essential The e-h pair distribution around single trajectories must be modeled for ULSI. (This overlaps biophysics, cells, DNA, etc.) Little issues (suggestions): Add energy categories, ionization, phonons, displacement, radiation, etc., (in G4Step?) for processes that wish to use them. Total deposited energy is not sufficient. Require as policy that all variables be accessible to derived classes unless doing so breaks functionality. Finally, (far out) Define membranes (zero-volume, surface-only objects with unique surface normal at each point) for parallel worlds. Not for material world. Purpose: Event metrology. 19