GEant4 Microdosimetry Analysis Tool - GEMAT

Fan Lei, Peter Truscott & Petteri NieminenSPENVIS/Geant4 Workshop, Leuven, Belgium05 October 2005

2

Contents

1 Background

2 GEMAT overview

– Geometry Builder

– Physics list

– Analysis manager

3 Application example

4 Further developments

5 Summary

3

Background

• Single event effects are data corruptions or failures induced in microelectronics by single particles

• These are a major factor limiting the reliability of future microelectronics. Susceptibilities to SEEs and the range of effects are on the increase

• Currently, best way to quantify device susceptibility to energetic particles is to use accelerator facilities - expensive, may not relate to operational conditions, and does not tell us about the physics processes

4

SEE Modelling Objectives

• Develop modelling capability to simulate high-energy interaction processes, charge production, and semiconductor device response

• In doing so:– Reduce the reliance on repeated recourse to experiments to

determine device susceptibility– Enable better understanding of dominant physical processes

driving observed effects• Provide an engineering tool to assist in cost-effective selection

of current/future components for aerospace and general safety-critical projects

• At QinetiQ SEE modelling activities are supported by the MoD/CRP and by ESA through the REAT, REAT-MS contracts.

5

Simulation of the Single Event Effect Processes

Need develop models for the complete process, these include

1. nuclear interaction

2. charge generation

3. charge collection

• This talk is addressing process 1 only

Gate

Electroncurrent

Funnel

Substrate

Depletion layer

Polysilicon plate

Incident particle

Recoil nucleus

6GEant4 Microdosimetry Analysis Tool - GEMATA Geant4 based application for microdosimetry analysis of

microelectronics

• Easy to use geometry builder

– Handle more complex volumes than regular parallelepiped

• Dedicated physics list

– Making use of the full G4 physics capability

• Build-in analysis modes

– PHS: SEU rates calculated based on experimental ion data

– Path-length: used with environment h-ion LET data

– …

7

GEMAT Overview

• It is a standard Geant4 application:– Geometry construction

– Primary particle generation

– Physics list

– Histogram/Analysis manager

RunGMARunAction

EventGMAEventAction

TrackingGMATrackingActionGMASteppingAction

Digits & Hits ProcessesGMAPhysicsListn

TrackGMAPrimaryGeneratorActionGMAGeneralParticleSource

GeometryGMAGeometryDescription

Particle

MaterialIntercoms

GMAAnalysisMessengerGMAGeometryMessengerGMAEventActionMessengerGMAGeneralParticleSourceMessenger

InterfacesGMAVisManager

Geant4

HistogramingGMAAnalysisManagerGMAHisto1DHisto1DVarVariableLengthPartitionCSVofstream

BinningQ4BinScheme

main

8

Geometry Constructor Options

• C++ coding

• Geometry/structure file from SILVACO device physics code

• Using the build-in geometry commands:

– Material definition

– Geometry definition

– Visualization attributes

9

Material Definition Commands

• /geometry/material/add

» /delete

» /list

List of commands:

Predefined material:

• There are 4 predefined materials

• New material can be added by given its name, element composition and density

e.g. /geometry/material/add SiOxide Si-O2 2.7

10

Geometry Construction Commands

• A layered geometry structure– Arbitrary number of layers

of different materials

• One layer is designated as the Contact Layer– Contact Volumes (CVs) can

be added

• One layer is designated as the Depleted Layer – Sensitive Volumes (SVs)

can be added Depleted regions non-depleted active or

inactive regions

x

y

z

Contacts

11

CV/DV Shapes

• Basic shapes

– Cylinder: 2 parameters

– Box: 2 parameters

– L-shape: 4 parameters

– U-shape: 4 parameters

• All can be tapered at top/bottom

• Position (x,y) in the layer

• Material & Vis. Attr.

Cylinder Rectangular Parallelepiped

'L' shape

'U' shape

12

Physics List

• G4LowEnergyEM

• G4HPNeutron

• G4Binary/G4Bertini

• G4BinaryLightIon

• G4Abrasion/G4Ablation

• G4RadioactiveDecay

• Layer dependent cut-offs

• Max step-size, max frac. Of energy loss

• Bias the C-S of a process

Primary Particle GeneratorG4GeneralParticleSource (GPS)

13

Analysis Manager

• Fluence, Pulse Height Spectrum (PHS) and Path-length

• Applied to selected sensitive volumes (SVs)

• Build-in histogram capability

– Wide choice of binning scheme, inc. arbitrary

– Output in CSV format

Coincidence analysis:

– Between up to 3 DVs

– Each volume can have its own threshold

14An application Example: 4 Mbit SRAMs• A large amount of beam test data available, from

heavy Ion to thermal neutrons

• Good knowledge of the device geometry

• Two types of simulations using

– Detailed geometry at cell level

– An array of simple cells

15GEMAT geometry for four-transistor cell,

forming part of a 4Mbit SRAM

Pink-outlined regions indicate sensitive volumes

16

1.E-18

1.E-17

1.E-16

1.E-15

0 10 20 30 40 50 60 70

Proton energy [MeV]S

EU

cro

ss s

ecti

on

[cm

2 /bit

]

Experiment data from Poivey

G4 Classical Cascade modelpredictionG4 Binary Cascade modelprediction

Proton SEU predictions for Samsung KM684002A 4Mbit SRAM

The energy-deposition spectrum from events in SVs integrated over a Weibull fit to LET data from heavy-ion tests

17

Neutron Data

1.E-16

1.E-15

1.E-14

1.E-13

1.E-12

0 100 200 300 400 500

Particle Energy (MeV)

SE

U-b

it X

-Se

ctio

n (

cm2)

G4 Neutron Simulation (Weibull 1)Neutron ExperimentalIRTS Neutron Simulation

SVs modelled as an array of simple 0.5x0.5m2 cells

18

MBU rates

1.E-20

1.E-19

1.E-18

1.E-17

1.E-16

1.E-15

1.E-14

1.E-13

1.E-12

0 100 200 300 400 500 600

Proton Energy (MeV)

SE

U C

ross

-Sec

tion

(cm

2)

Single Double Triple

1.E-19

1.E-18

1.E-17

1.E-16

1.E-15

1.E-14

1.E-13

1.E-12

0 100 200 300 400 500 600

Neutron Energy (MeV)

SE

U C

ross

-Sec

tion

(cm

2)

Single Double Triple

Proton Neutron

19

Further DevelopmentsWithin the REAT-MS project

– Porting to the GRAS simulation framework

– Geometry:• GDML• CVs/DVs in any layer

– Physics: improved biasing– Analysis:

• AIDA • SEU rate calculations

– Integration into SPENVIS

In the longer term:

– Ion track models

– Charge generation process

• Phonons and plasmason

– Low energy Ion nuclear interaction (< 100 MeV/nucl.)

– More SEU algorithms and models

– …

20

Summary

• Modelling is required to reduce the reliance of SEE study on beam experiments and for understanding of the underlying physics processes

• Much of the physics to perform detailed single-event simulations are in place in Geant4. GEMAT provided the accessibility to this powerful toolkit

• It also provide easy to use methods for definition of 3D microdosimetry geometry representing semiconductor, and incident particles (spectrum, angular distribution)

• GEMAT has already been successfully used in a wide range of SEE analysis

• New developments planned in the REAT-MS project will further improve its capability and usability, making it available in the SPENVIS system