Radiation Damage Mechanisms
in CCD Imagers
Paul W. Marshall1,2 and Cheryl J. Marshall 1
1. NASA-GSFC
2. Consultant
STScI Workshop on HST CCD Detector CTE
January 31- February 1, 2000
pwmarshall@aol.com: 804-376-3402
cheryl.marshall@gsfc.nasa.gov: 804-376-3402
Outline
• Overview of Radiation Effects in CCDs
• Proton-Induced Displacement Damage Mechanisms
• Interaction Physics and Nonionizing Energy Loss
• Shielding Issues
• Microdosimetry Issues
• On-Orbit Prediction Methods and Tools for CTE and Jd
• Ingredients: Proton Spectrum at Device Location in Satellite,
Proton Measurements of Important Effects at Specific Energies,
Energy Dependence of Displacement Damage Effects
• Fidelity with Mission Requirements and Sources of Uncertainty
Section 1
• Overview of Radiation Effects in CCDs
• Ionization Transients
• Total Ionizing Dose
• Displacement Damage Introduction
Proton Direct Ionization and SEEs
• CCD Transient Ionization Events
• Proton Ionization Tracks Release Hundreds to
Thousands of Charge Pairs per Micron Pathlength
• Transient Signals on Star Tracker CCDs May Be
Suppressed with Temporal (Kalman) Filtering
− Rates on CCD Can Be > 103 /sec.
T.S. Lomheim, et al., “Imaging Charged-Coupeld Device (CCD) Transient
Response to 17 and 50 MeV Proton and Heavy Ion Irradiation,”
IEEE TNS, Vol. 37, No. 6, 1990.
SOHO/LASCO CCD During Solar Particle Event
Proton Stopping Power and Range in Si
Proton Dose Deposition in Silicon
ISSA Depth - Dose Relation
Total Ionizing Dose Effects in CCDs
• Flatband Shifts that Shift the Bias Windows for Optimal
CCD Operation and Degrade Output Amplifier Performance
• Increase in Surface Dark Current
• Hardening Solutions have been Developed:
• Robust Oxides and/or Voltage Adjustments
• Multi-Phased Pinning : Operation with Si Surface Inverted to
Keep Surface States Filled to Minimize Dark Currents
• Reduced Operating Temperatures to Minimize Dark Currents
Section 2
• Proton-Induced Displacement Damage Mechanisms
• Interaction Physics for Coulombic and Nuclear Scattering
• Defect Formation
• Electrical Properties of Defects
• Defect Behavior in CCDs
Primary Concern is Displacement Damage
• In general, CCD Degradation Due to Cobalt 60 TID Levels
Less than 10-20 krad(Si) are Manageable
• In contrast, CTE Losses from Displacement Damage Can
be Significant for Proton Exposures at the 1 krad(Si) Level.
• Hardening Solutions for Displacement Damage are not
Well-Developed
• Protons are Difficult to Shield
• Hardening Techniques Generally Provide Incremental
Improvements for Special Cases
• Use of p-channel CCDs is Being Investigated
Evolution of Initial Vacancy-Interstitial Pairs
to Stable Defects
EXITING
PARTICLE
INCIDENT
PARTICLE
STABLE
DEFECT
Interstitial
Vacancy
Dopant or Impurity Atom
Displacement Damage Processes in Si
6-10 MeV
Log N
FREE
DEFECTS,
Coulomb
> 20 MeV
SINGLE
CASCADE,
Nuclear Elastic
PROTON ENERGY
MANY SUBCASCADES,
Nuclear Reactions
RECOIL ENERGY
1-2 keV
12-20 keV
Displacement Damage Formation
•PKA spectrum and defect formation
–PKA transport and subsequent collisions
–Initial quench (~90% of vacancies recombine)
–Frenkel pair formation
–Diffusion of vacancies and interstitial atoms
–Defect complex formation (e.g., E-center)
–Defect complex stability (annealing)
•Quantitative determination of concentrations of
specific defects must rely on semi-empirical estimates
–Depends on type and energy of radiation, and material
(dopant and impurity concentrations)
Electrical Effects of Radiation-Induced Defects
Conduction Band
EDONOR
EC
ET
EV
Valence Band
Generation
Trapping
Recombination
Compensation
Tunneling
Trap Emission Time Constant (τe)
 Et 
g
τ =
exp 
 kT 
e σ ν N
 
n t c
where Et = trap energy, T = temperature, k = Boltzmann’s constant,
σn = electron capture cross section, νt = electron thermal velocity,
Nc = conduction band density of states, and g = level degeneracy
• Imagine a Charge Packet Encountering an Empty Pixel with Traps.
Since the Electron Capture Time Constants are Very Short
Relative to the Clock Period, Electrons Will Fill the Empty Traps.
• The Probability that a Trapped Electron Will Not
Be Emitted During a Given Time is Given by:
 − Time 

exp
 τ

e 

V3V1
Direction of
charge transfer
oxide
p-epi layer
V2
V
V3 1
V
V1 2
p-substrate
channel
stops
buried
n-channel
CTE/pixel =
1 -
# Electrons Lost
# Transfers x Signal Size
CCD Charge
Transfer & Readout
Processes
Channel Stops
V1
V2
V3
Parallel
Transfers
in the
Imaging
Region
Output
Serial
Readout
Register
V1V2V3
CTI Using Shockley-Read-Hall Theory
V
T
T
s
t
CTI PIXEL = (
* N t )(exp ( − ))(1 − exp ( − s ))
N
τ
τ
s
e
e
Vs = signal volume, N s = # signal electrons, Nt = # traps,
Tt = clock period, Ts = time between x-ray events, τe = trap emission time
A Charge Packet Encounters An Empty Pixel....
• # Trapped Charges Relative to Size of Charge Packet
• Probability a Charge Will Remain Trapped
• Probability a Charge Will Be Emitted Before the Next Packet
Arrives & Trap will be Empty
Charge Transfer Inefficiency vs. Temperature
•SRH theory help assess
the temperature effects
of the CTE loss from
various trap populations
and energy levels in a
frequency and scene
dependent manner. The
two curves for the Ecenter show the effect of
x-ray signal density. If
background signal is
present or if pixels with
signal are close together,
a fat zero effect results.
C.J. Dale, et al., “Displacement Damage Effects in Mixed Particle Environments
for Shielded Spacecraft CCDs”, IEEE TNS, Vol. 40, No. 6, pp. 1628-1637, 1993.
Section 3
• On-Orbit Prediction Methods and Tools for CTE and Jd
• Ingredients: Proton Spectrum at Device Location in Satellite,
Proton Measurements of Important Effects at Specific Energies,
Energy Dependence of Displacement Damage Effects
• Fidelity with Mission Requirements and Sources of Uncertainty
• Microdosimetry Considerations for Dark Current and
CTE Loss
∆ DEVICE PROPERTY
On-Orbit Predictions: Device Degradation
E1
E2
Damage
Factor
=
∆ Device Property
∆ Fluence
PROTON FLUENCE
• Application Specific Measurements, e.g. CTE or Average Dark Current
• Consideration of Possible TID Effects
DEVICE
PARAMETER
PROTON ENERGY (MeV)
Protons / cm2 • day • MeV
DAMAGE FACTOR
NIEL (MeV • cm2/g)
On-Orbit Performance Predictions
TIME IN ORBIT
PROTON ENERGY (MeV)
Proton Energy Loss Mechanisms in Si:
Ionization (LET) and Displacements (NIEL)
C.J. Dale et al., “A Comparison of Monte Carlo and Analytic Treatments
of Displacement Damage in Microvolumes, IEEE TNS, Vol. 41, 1994.
Comparison of CTE Damage Factor Energy
Dependence with NIEL Calculation
•
NIEL (MeV cm2•g-1)
Leicester
JPL
10-1
10-2
10-11
10-12
NIEL
10-13
10-3
10-14
10-4
10-1
CTE DAMAGE FACTOR
(∆CTE cm2/Proton)
100
CTE D.F.
NIEL
=
1.2x10-11 ∆ CTE
MeV/g
100
101
102
103
PROTON ENERGY (MeV)
C.J. Dale, et al., “Displacement Damage Effects in Mixed Particle Environments
for Shielded Spacecraft CCDs”, IEEE TNS, Vol. 40, No. 6, pp. 1628-1637, 1993.
On-Orbit Performance Predictions
• Experiments Show That, to First Order, NIEL Describes the
Energy Dependence of Device Degradation
• Non-Ionizing Energy Loss Rate (NIEL) Plays the Same Role in
Displacement Damage Effects as LET Does for TID Effects
• Displacement Damage “Dose” = DDD = NIEL(E)[dΦ/dE]dE
• Units are MeV/g(material)
• Φ(Etest ) = DDD / NIEL(Etest)
• Choose Test Energy Wisely
• Experimental Damage Functions May Also Be Used
On-Orbit Prediction of CTE vs. Shield Thickness
•Thick Shields Required: ~ Half Damage Due to >100 MeV Protons!
Shielding Solution for Improved CTE in CCDs
• High Z Shields Desired
• Nuclear Reactions w/
Incoming Protons Create
Secondary Particles
• Total NIEL is Sum of
Primary Proton NIEL &
Secondary Neutron NIEL
• Significant Neutron
Damage from Thick High
Z Shields
C.J. Dale, et al., “Displacement Damage Effects in Mixed Particle Environments for
Shielded Spacecraft CCDs”, IEEE TNS, Vol. 40, No. 6, pp. 1628-1637, 1993.
Proton-Induced Dark Current Increases
CHANGE IN DARK CURRENT(nA•cm-2)
• Average Dark Current Increases Due to Carrier Generation in the Bulk
14.0
12.0
12 MeV p +
22 MeV p +
• Dark Current
Nonuniformity
Important !
10.0
63 MeV p +
8.0
(Spatial &
Temporal)
6.0
4.0
2.0
0.0
0.0
1.0
2.0
3.0
4.0
5.0
FLUENCE (x 10 11 cm-2)
6.0
7.0
C.J. Dale, et al., “The Generation Lifetime Damage Constant and its Variance,” IEEE TNS, 1989
Calculated Proton Recoil Parameters for Silicon
Calculated Proton Damage Distributions
•Proton damage occurs
due to a combination of
very frequent elastic
and less frequent
nuclear elastic and
inelastic processes. In
this fluence regime, all
pixels have many
elastic events but few
of the more energetic
inelastic events. Both
processes are random
in nature and Poisson
statistics apply.
Calculated Proton Damage Distributions
•Calculated damage
distributions for a
given proton energy,
fluence, and pixel
geometry are
constructed by
convolving the
distributions for
elastic and inelastic
damage. The
composite
distribution follows
from Poisson
weighting and
superposition.
Comparison of Measured Dark Current Histogram
with Calculated Proton Damage
•By normalizing the mean of the
calculated damage energy
distribution with the mean of the
measured dark current histogram,
the shapes are compared. There
are no free parameters, and
agreement in the skewed high
energy tail indicates the
applicability of the model to the
problem of dark spikes.
Extension of Calculated Damage to Other Fluences
•The model is readily
extended to predict
dark current
histograms at other
fluences and at other
proton energies. Note
the Gaussian nature
as the average
number of inelastic
events increases.
At higher proton
energies the role of
the inelastic recoils
increases
dramatically.
Pixel-to-Pixel Dark Current Nonuniformities
• Nonuniformity
Can Be Larger
Concern
• Statistical Nature
of Collision
Kinematics
• High Energy
Tails Due to
Multiple Nuclear
Reactions/Pixel
P.W. Marshall, et al., “Proton-Induced Displacement Damage Distributions
and Extremes in Silicon Microvolumes,” IEEE Trans. Nucl. Sci., NS-37, Vol. 6, 1990.
Calculated Proton Recoil Parameters for Silicon
•Extension of the
model to lower
fluences reveals the
bimodal distribution
for which most pixels
have only elastic
damage, but those
with inelastic damage
have much more
damage than their
neighbors. Again, this
effect is more
pronounced at higher
proton energies
Microdosimetry Considerations for CTE Loss
<1.2 µm Overlayer
•Extension of the preceding analysis to
the small channel volumes associated
with CTE effects are difficult to assess
analytically because of nonequilibrium
recoil spectrum issues near the surface
and also because most energetic recoil
atoms will cross the boundries of the
channel.
•The observed scatter in the Fe-55 xray based stacked line trace indicates
that column-to-column variations in
CTE loss are significant. Further
analysis is in progress.
0.15 - 1 µm CTE Volume
Incoming Protons
C.J. Dale, L. Chen, P.J. McNulty, P.W. Marshall, and E.A Burke, “A Comparison of Monte
Carlo and Analytic Treatments of Displacement Damage in Microvolumes,” IEEE TNS, 1994.
INELASTIC DAMAGE ENERGY
DEPOSITED PER UNIT FLUENCE
(MeV•cm2)
Approach to Displacement Damage Equilibrium
8x10-14
Sensitive Volume
11 x 7 x 0.15 µm3
20 MeV Protons
• Recoil Equilibrium
Reached After
Several Microns
6x10-14
63 MeV Protons
4x10-14
150 MeV Protons
2x10-14
0
0
2
4
6
8
OVERLAYER THICKNESS (µm)
C.J. Dale, L. Chen, P.J. McNulty, P.W. Marshall, and E.A Burke, “A Comparison of Monte
Carlo and Analytic Treatments of Displacement Damage in Microvolumes,” IEEE TNS, 1994.
Dark Current Density (pA/cm2)
Temporal Dark Current Fluctuations in a Pixel
EEV CCD,
200
10 MeV p +
-10 °C
150
100
50
EEV CCD05, 10 MeV Protons
0
1
2
Time (Hours)
3
4
I.H. Hopkins and G.R. Hopkinson, “Random Telegraph Signals from
Proton-Irradiated CCDs,” IEEE TNS, Vol. 40, No. 6, pp. 1567-1574, 1993.
Sources of Uncertainty in Prediction of CTE Changes
• Environment Considerations
• Mission proton exposure (factor of 2 uncertainty)
• Fidelity of shielding model
• Neutron production and effects
• Experimental Variables
•
•
•
•
Device variability
Facility dosimetry
CTE measurement fidelity (Fe-55, EPER, FPR, etc.)
Microdosimetry considerations
• Prediction Models
• NIEL applicability, particle and energy dependence
• Cumulative Uncertainty May be as Much as 2x to 3x
Summary
• Dark current from displacement damage can be
managed with cooling, but CTE is more tricky
• Modeling can predict general behavior
• Physical process governing damage are mostly
understood though precise predictions not easy
• Shielding helps but cannot solve the problem
• Neutron effects must be considered
• On-orbit predictions are possible with limited test
data and damage correlation with NIEL
• Requires fidelity with temperature, readout scheme,
scene, signal size, etc.
• Microdosimetry of damage should be considered